Artificial Intelligence Nanodegree

Computer Vision Capstone

Project: Facial Keypoint Detection


Welcome to the final Computer Vision project in the Artificial Intelligence Nanodegree program!

In this project, you’ll combine your knowledge of computer vision techniques and deep learning to build and end-to-end facial keypoint recognition system! Facial keypoints include points around the eyes, nose, and mouth on any face and are used in many applications, from facial tracking to emotion recognition.

There are three main parts to this project:

Part 1 : Investigating OpenCV, pre-processing, and face detection

Part 2 : Training a Convolutional Neural Network (CNN) to detect facial keypoints

Part 3 : Putting parts 1 and 2 together to identify facial keypoints on any image!


*Here's what you need to know to complete the project:

  1. In this notebook, some template code has already been provided for you, and you will need to implement additional functionality to successfully complete this project. You will not need to modify the included code beyond what is requested.

    a. Sections that begin with '(IMPLEMENTATION)' in the header indicate that the following block of code will require additional functionality which you must provide. Instructions will be provided for each section, and the specifics of the implementation are marked in the code block with a 'TODO' statement. Please be sure to read the instructions carefully!

  1. In addition to implementing code, there will be questions that you must answer which relate to the project and your implementation.

    a. Each section where you will answer a question is preceded by a 'Question X' header.

    b. Carefully read each question and provide thorough answers in the following text boxes that begin with 'Answer:'.

Note: Code and Markdown cells can be executed using the Shift + Enter keyboard shortcut. Markdown cells can be edited by double-clicking the cell to enter edit mode.

The rubric contains optional suggestions for enhancing the project beyond the minimum requirements. If you decide to pursue the "(Optional)" sections, you should include the code in this IPython notebook.

Your project submission will be evaluated based on your answers to each of the questions and the code implementations you provide.

Steps to Complete the Project

Each part of the notebook is further broken down into separate steps. Feel free to use the links below to navigate the notebook.

In this project you will get to explore a few of the many computer vision algorithms built into the OpenCV library. This expansive computer vision library is now almost 20 years old and still growing!

The project itself is broken down into three large parts, then even further into separate steps. Make sure to read through each step, and complete any sections that begin with '(IMPLEMENTATION)' in the header; these implementation sections may contain multiple TODOs that will be marked in code. For convenience, we provide links to each of these steps below.

Part 1 : Investigating OpenCV, pre-processing, and face detection

  • Step 0: Detect Faces Using a Haar Cascade Classifier
  • Step 1: Add Eye Detection
  • Step 2: De-noise an Image for Better Face Detection
  • Step 3: Blur an Image and Perform Edge Detection
  • Step 4: Automatically Hide the Identity of an Individual

Part 2 : Training a Convolutional Neural Network (CNN) to detect facial keypoints

  • Step 5: Create a CNN to Recognize Facial Keypoints
  • Step 6: Compile and Train the Model
  • Step 7: Visualize the Loss and Answer Questions

Part 3 : Putting parts 1 and 2 together to identify facial keypoints on any image!

  • Step 8: Build a Robust Facial Keypoints Detector (Complete the CV Pipeline)

Step 0: Detect Faces Using a Haar Cascade Classifier

Have you ever wondered how Facebook automatically tags images with your friends' faces? Or how high-end cameras automatically find and focus on a certain person's face? Applications like these depend heavily on the machine learning task known as face detection - which is the task of automatically finding faces in images containing people.

At its root face detection is a classification problem - that is a problem of distinguishing between distinct classes of things. With face detection these distinct classes are 1) images of human faces and 2) everything else.

We use OpenCV's implementation of Haar feature-based cascade classifiers to detect human faces in images. OpenCV provides many pre-trained face detectors, stored as XML files on github. We have downloaded one of these detectors and stored it in the detector_architectures directory.

Import Resources

In the next python cell, we load in the required libraries for this section of the project.

In [1]:
# Import required libraries for this section

%matplotlib inline

import numpy as np
import matplotlib.pyplot as plt
import math
import cv2                     # OpenCV library for computer vision
from PIL import Image
import time 

Next, we load in and display a test image for performing face detection.

Note: by default OpenCV assumes the ordering of our image's color channels are Blue, then Green, then Red. This is slightly out of order with most image types we'll use in these experiments, whose color channels are ordered Red, then Green, then Blue. In order to switch the Blue and Red channels of our test image around we will use OpenCV's cvtColor function, which you can read more about by checking out some of its documentation located here. This is a general utility function that can do other transformations too like converting a color image to grayscale, and transforming a standard color image to HSV color space.

In [2]:
# Load in color image for face detection
image = cv2.imread('images/test_image_1.jpg')

# Convert the image to RGB colorspace
image = cv2.cvtColor(image, cv2.COLOR_BGR2RGB)

# Plot our image using subplots to specify a size and title
fig = plt.figure(figsize = (8,8))
ax1 = fig.add_subplot(111)
ax1.set_xticks([])
ax1.set_yticks([])

ax1.set_title('Original Image')
ax1.imshow(image)
Out[2]:
<matplotlib.image.AxesImage at 0x7ff6b07d8400>

There are a lot of people - and faces - in this picture. 13 faces to be exact! In the next code cell, we demonstrate how to use a Haar Cascade classifier to detect all the faces in this test image.

This face detector uses information about patterns of intensity in an image to reliably detect faces under varying light conditions. So, to use this face detector, we'll first convert the image from color to grayscale.

Then, we load in the fully trained architecture of the face detector -- found in the file haarcascade_frontalface_default.xml - and use it on our image to find faces!

To learn more about the parameters of the detector see this post.

In [3]:
# Convert the RGB  image to grayscale
gray = cv2.cvtColor(image, cv2.COLOR_RGB2GRAY)

# Extract the pre-trained face detector from an xml file
face_cascade = cv2.CascadeClassifier('detector_architectures/haarcascade_frontalface_default.xml')

# Detect the faces in image
faces = face_cascade.detectMultiScale(gray, 4, 6)

# Print the number of faces detected in the image
print('Number of faces detected:', len(faces))

# Make a copy of the orginal image to draw face detections on
image_with_detections = np.copy(image)

# Get the bounding box for each detected face
for (x,y,w,h) in faces:
    # Add a red bounding box to the detections image
    cv2.rectangle(image_with_detections, (x,y), (x+w,y+h), (255,0,0), 3)
    

# Display the image with the detections
fig = plt.figure(figsize = (8,8))
ax1 = fig.add_subplot(111)
ax1.set_xticks([])
ax1.set_yticks([])

ax1.set_title('Image with Face Detections')
ax1.imshow(image_with_detections)
Number of faces detected: 13
Out[3]:
<matplotlib.image.AxesImage at 0x7ff6b0774ef0>

In the above code, faces is a numpy array of detected faces, where each row corresponds to a detected face. Each detected face is a 1D array with four entries that specifies the bounding box of the detected face. The first two entries in the array (extracted in the above code as x and y) specify the horizontal and vertical positions of the top left corner of the bounding box. The last two entries in the array (extracted here as w and h) specify the width and height of the box.


Step 1: Add Eye Detections

There are other pre-trained detectors available that use a Haar Cascade Classifier - including full human body detectors, license plate detectors, and more. A full list of the pre-trained architectures can be found here.

To test your eye detector, we'll first read in a new test image with just a single face.

In [4]:
# Load in color image for face detection
image = cv2.imread('images/james.jpg')

# Convert the image to RGB colorspace
image = cv2.cvtColor(image, cv2.COLOR_BGR2RGB)

# Plot the RGB image
fig = plt.figure(figsize = (6,6))
ax1 = fig.add_subplot(111)
ax1.set_xticks([])
ax1.set_yticks([])

ax1.set_title('Original Image')
ax1.imshow(image)
Out[4]:
<matplotlib.image.AxesImage at 0x7ff64bc73f98>

Notice that even though the image is a black and white image, we have read it in as a color image and so it will still need to be converted to grayscale in order to perform the most accurate face detection.

So, the next steps will be to convert this image to grayscale, then load OpenCV's face detector and run it with parameters that detect this face accurately.

In [5]:
# Convert the RGB  image to grayscale
gray = cv2.cvtColor(image, cv2.COLOR_RGB2GRAY)

# Extract the pre-trained face detector from an xml file
face_cascade = cv2.CascadeClassifier('detector_architectures/haarcascade_frontalface_default.xml')

# Detect the faces in image
faces = face_cascade.detectMultiScale(gray, 1.25, 6)

# Print the number of faces detected in the image
print('Number of faces detected:', len(faces))

# Make a copy of the orginal image to draw face detections on
image_with_detections = np.copy(image)

# Get the bounding box for each detected face
for (x,y,w,h) in faces:
    # Add a red bounding box to the detections image
    cv2.rectangle(image_with_detections, (x,y), (x+w,y+h), (255,0,0), 3)
    

# Display the image with the detections
fig = plt.figure(figsize = (6,6))
ax1 = fig.add_subplot(111)
ax1.set_xticks([])
ax1.set_yticks([])

ax1.set_title('Image with Face Detection')
ax1.imshow(image_with_detections)
Number of faces detected: 1
Out[5]:
<matplotlib.image.AxesImage at 0x7ff64bbdca20>

(IMPLEMENTATION) Add an eye detector to the current face detection setup.

A Haar-cascade eye detector can be included in the same way that the face detector was and, in this first task, it will be your job to do just this.

To set up an eye detector, use the stored parameters of the eye cascade detector, called haarcascade_eye.xml, located in the detector_architectures subdirectory. In the next code cell, create your eye detector and store its detections.

A few notes before you get started:

First, make sure to give your loaded eye detector the variable name

eye_cascade

and give the list of eye regions you detect the variable name

eyes

Second, since we've already run the face detector over this image, you should only search for eyes within the rectangular face regions detected in faces. This will minimize false detections.

Lastly, once you've run your eye detector over the facial detection region, you should display the RGB image with both the face detection boxes (in red) and your eye detections (in green) to verify that everything works as expected.

In [6]:
# Make a copy of the original image to plot rectangle detections
image_with_detections = np.copy(image)   

# Loop over the detections and draw their corresponding face detection boxes
for (x,y,w,h) in faces:
    cv2.rectangle(image_with_detections, (x,y), (x+w,y+h),(255,0,0), 3)  
    
# Do not change the code above this comment!

    
## TODO: Add eye detection, using haarcascade_eye.xml, to the current face detector algorithm
eye_cascade = cv2.CascadeClassifier('detector_architectures/haarcascade_eye.xml')

## Iterate faces and detect eyes
eyes = []
for (x,y,w,h) in faces:
    # Detect the eyes in each face
    eyes_tmp = eye_cascade.detectMultiScale(gray[y:(y+h), x:(x+w)], 1.2, 3)
    # add x and y to the eyes coordinates
    eyes_tmp[:,:2] = eyes_tmp[:,:2] + np.array([x, y])
    eyes.append(eyes_tmp)

# convert eyes coordinaes to numpy array
eyes = np.squeeze(np.array(eyes), axis=0)
     
## TODO: Loop over the eye detections and draw their corresponding boxes in green on image_with_detections
for (x,y,w,h) in eyes:
    # Add a green bounding box to the detections image
    cv2.rectangle(image_with_detections, (x,y), (x+w,y+h), (0,255,0), 2)

# Plot the image with both faces and eyes detected
fig = plt.figure(figsize = (6,6))
ax1 = fig.add_subplot(111)
ax1.set_xticks([])
ax1.set_yticks([])

ax1.set_title('Image with Face and Eye Detection')
ax1.imshow(image_with_detections)
Out[6]:
<matplotlib.image.AxesImage at 0x7ff64bb5e978>

(Optional) Add face and eye detection to your laptop camera

It's time to kick it up a notch, and add face and eye detection to your laptop's camera! Afterwards, you'll be able to show off your creation like in the gif shown below - made with a completed version of the code!

<img src="images/laptop_face_detector_example.gif" width=400 height=300/>

Notice that not all of the detections here are perfect - and your result need not be perfect either. You should spend a small amount of time tuning the parameters of your detectors to get reasonable results, but don't hold out for perfection. If we wanted perfection we'd need to spend a ton of time tuning the parameters of each detector, cleaning up the input image frames, etc. You can think of this as more of a rapid prototype.

The next cell contains code for a wrapper function called laptop_camera_face_eye_detector that, when called, will activate your laptop's camera. You will place the relevant face and eye detection code in this wrapper function to implement face/eye detection and mark those detections on each image frame that your camera captures.

Before adding anything to the function, you can run it to get an idea of how it works - a small window should pop up showing you the live feed from your camera; you can press any key to close this window.

Note: Mac users may find that activating this function kills the kernel of their notebook every once in a while. If this happens to you, just restart your notebook's kernel, activate cell(s) containing any crucial import statements, and you'll be good to go!

In [7]:
### Add face and eye detection to this laptop camera function 
# Make sure to draw out all faces/eyes found in each frame on the shown video feed

import cv2
import time 

# wrapper function for face/eye detection with your laptop camera
def laptop_camera_go():
    # Create instance of video capturer
    cv2.namedWindow("face detection activated")
    vc = cv2.VideoCapture(0)

    # Try to get the first frame
    if vc.isOpened(): 
        rval, frame = vc.read()
    else:
        rval = False
    
    # Extract the pre-trained face and eyes detector from the xml files
    face_cascade = cv2.CascadeClassifier('detector_architectures/haarcascade_frontalface_default.xml')
    eye_cascade = cv2.CascadeClassifier('detector_architectures/haarcascade_eye.xml')
    
    # Keep the video stream open
    while rval:
        # Convert the RGB  image to grayscale
        gray = cv2.cvtColor(frame, cv2.COLOR_RGB2GRAY)
        
        # Detect the faces in image
        faces = face_cascade.detectMultiScale(gray, 1.25, 5)
        
        # Make a copy of the orginal image to draw face detections on
        frame_with_detections = np.copy(frame)
        
        ## Iterate faces and draw bounding box, detect eyes and draw bounding boxes for eyes in green
        eyes = []
        for (x,y,w,h) in faces:
            # Add a red bounding box to the detections image
            cv2.rectangle(frame_with_detections, (x,y), (x+w,y+h), (255,0,0), 3)
            # Detect the eyes in each face
            eyes = eye_cascade.detectMultiScale(gray[y:(y+h), x:(x+w)], 1.2, 2)
            for (xe,ye,we,he) in eyes:
                # Add a green bounding box to the detections image
                cv2.rectangle(frame_with_detections, (x+xe,y+ye), (x+xe+we,y+ye+he), (0,255,0), 2)
        
        # Plot the image from camera with all the face and eye detections marked
        cv2.imshow("face detection activated", frame_with_detections)
        
        # Exit functionality - press any key to exit laptop video
        key = cv2.waitKey(2)
        if key < 255: # Exit by pressing any key
            # Destroy windows 
            cv2.destroyAllWindows()
            
            # Make sure window closes on OSx
            for i in range (1,5):
                cv2.waitKey(1)
            return
        
        # Read next frame
        time.sleep(0.5)             # control framerate for computation - default 20 frames per sec
        rval, frame = vc.read()    
In [ ]:
# Call the laptop camera face/eye detector function above
laptop_camera_go()

Step 2: De-noise an Image for Better Face Detection

Image quality is an important aspect of any computer vision task. Typically, when creating a set of images to train a deep learning network, significant care is taken to ensure that training images are free of visual noise or artifacts that hinder object detection. While computer vision algorithms - like a face detector - are typically trained on 'nice' data such as this, new test data doesn't always look so nice!

When applying a trained computer vision algorithm to a new piece of test data one often cleans it up first before feeding it in. This sort of cleaning - referred to as pre-processing - can include a number of cleaning phases like blurring, de-noising, color transformations, etc., and many of these tasks can be accomplished using OpenCV.

In this short subsection we explore OpenCV's noise-removal functionality to see how we can clean up a noisy image, which we then feed into our trained face detector.

Create a noisy image to work with

In the next cell, we create an artificial noisy version of the previous multi-face image. This is a little exaggerated - we don't typically get images that are this noisy - but image noise, or 'grainy-ness' in a digitial image - is a fairly common phenomenon.

In [9]:
# Load in the multi-face test image again
image = cv2.imread('images/test_image_1.jpg')

# Convert the image copy to RGB colorspace
image = cv2.cvtColor(image, cv2.COLOR_BGR2RGB)

# Make an array copy of this image
image_with_noise = np.asarray(image)

# Create noise - here we add noise sampled randomly from a Gaussian distribution: a common model for noise
noise_level = 40
noise = np.random.randn(image.shape[0],image.shape[1],image.shape[2])*noise_level

# Add this noise to the array image copy
image_with_noise = image_with_noise + noise

# Convert back to uint8 format
image_with_noise = np.asarray([np.uint8(np.clip(i,0,255)) for i in image_with_noise])

# Plot our noisy image!
fig = plt.figure(figsize = (8,8))
ax1 = fig.add_subplot(111)
ax1.set_xticks([])
ax1.set_yticks([])

ax1.set_title('Noisy Image')
ax1.imshow(image_with_noise)
Out[9]:
<matplotlib.image.AxesImage at 0x7ff64ba43748>

In the context of face detection, the problem with an image like this is that - due to noise - we may miss some faces or get false detections.

In the next cell we apply the same trained OpenCV detector with the same settings as before, to see what sort of detections we get.

In [10]:
# Convert the RGB  image to grayscale
gray_noise = cv2.cvtColor(image_with_noise, cv2.COLOR_RGB2GRAY)

# Extract the pre-trained face detector from an xml file
face_cascade = cv2.CascadeClassifier('detector_architectures/haarcascade_frontalface_default.xml')

# Detect the faces in image
faces = face_cascade.detectMultiScale(gray_noise, 4, 6)

# Print the number of faces detected in the image
print('Number of faces detected:', len(faces))

# Make a copy of the orginal image to draw face detections on
image_with_detections = np.copy(image_with_noise)

# Get the bounding box for each detected face
for (x,y,w,h) in faces:
    # Add a red bounding box to the detections image
    cv2.rectangle(image_with_detections, (x,y), (x+w,y+h), (255,0,0), 3)
    

# Display the image with the detections
fig = plt.figure(figsize = (8,8))
ax1 = fig.add_subplot(111)
ax1.set_xticks([])
ax1.set_yticks([])

ax1.set_title('Noisy Image with Face Detections')
ax1.imshow(image_with_detections)
Number of faces detected: 13
Out[10]:
<matplotlib.image.AxesImage at 0x7ff64ba21e10>

With this added noise we now miss one of the faces!

(IMPLEMENTATION) De-noise this image for better face detection

Time to get your hands dirty: using OpenCV's built in color image de-noising functionality called fastNlMeansDenoisingColored - de-noise this image enough so that all the faces in the image are properly detected. Once you have cleaned the image in the next cell, use the cell that follows to run our trained face detector over the cleaned image to check out its detections.

You can find its official documentation here and a useful example here.

Note: you can keep all parameters except photo_render fixed as shown in the second link above. Play around with the value of this parameter - see how it affects the resulting cleaned image.

In [11]:
## TODO: Use OpenCV's built in color image de-noising function to clean up our noisy image!

# your final de-noised image (should be RGB)
denoised_image = cv2.fastNlMeansDenoisingColored(image_with_noise,None,20,20,7,21)
In [12]:
## TODO: Run the face detector on the de-noised image to improve your detections and display the result
# Convert the RGB  image to grayscale
gray_denoised_image = cv2.cvtColor(denoised_image, cv2.COLOR_RGB2GRAY)

# Detect the faces in image
faces = face_cascade.detectMultiScale(gray_denoised_image, 4, 6)

# Print the number of faces detected in the image
print('Number of faces detected:', len(faces))

# Make a copy of the orginal image to draw face detections on
denoised_image_with_detections = np.copy(denoised_image)

# Get the bounding box for each detected face
for (x,y,w,h) in faces:
    # Add a red bounding box to the detections image
    cv2.rectangle(denoised_image_with_detections, (x,y), (x+w,y+h), (255,0,0), 3)
    

# Display the image with the detections
fig = plt.figure(figsize = (8,8))
ax1 = fig.add_subplot(111)
ax1.set_xticks([])
ax1.set_yticks([])

ax1.set_title('Noisy Image with Face Detections')
ax1.imshow(denoised_image_with_detections)
Number of faces detected: 12
Out[12]:
<matplotlib.image.AxesImage at 0x7ff64b98b748>

Step 3: Blur an Image and Perform Edge Detection

Now that we have developed a simple pipeline for detecting faces using OpenCV - let's start playing around with a few fun things we can do with all those detected faces!

Importance of Blur in Edge Detection

Edge detection is a concept that pops up almost everywhere in computer vision applications, as edge-based features (as well as features built on top of edges) are often some of the best features for e.g., object detection and recognition problems.

Edge detection is a dimension reduction technique - by keeping only the edges of an image we get to throw away a lot of non-discriminating information. And typically the most useful kind of edge-detection is one that preserves only the important, global structures (ignoring local structures that aren't very discriminative). So removing local structures / retaining global structures is a crucial pre-processing step to performing edge detection in an image, and blurring can do just that.

Below is an animated gif showing the result of an edge-detected cat taken from Wikipedia, where the image is gradually blurred more and more prior to edge detection. When the animation begins you can't quite make out what it's a picture of, but as the animation evolves and local structures are removed via blurring the cat becomes visible in the edge-detected image.

<img src="images/Edge_Image.gif" width=400 height=300/>

Edge detection is a convolution performed on the image itself, and you can read about Canny edge detection on this OpenCV documentation page.

Canny edge detection

In the cell below we load in a test image, then apply Canny edge detection on it. The original image is shown on the left panel of the figure, while the edge-detected version of the image is shown on the right. Notice how the result looks very busy - there are too many little details preserved in the image before it is sent to the edge detector. When applied in computer vision applications, edge detection should preserve global structure; doing away with local structures that don't help describe what objects are in the image.

In [13]:
# Load in the image
image = cv2.imread('images/fawzia.jpg')

# Convert to RGB colorspace
image = cv2.cvtColor(image, cv2.COLOR_BGR2RGB)

# Convert to grayscale
gray = cv2.cvtColor(image, cv2.COLOR_RGB2GRAY)  

# Perform Canny edge detection
edges = cv2.Canny(gray,100,200)

# Dilate the image to amplify edges
edges = cv2.dilate(edges, None)

# Plot the RGB and edge-detected image
fig = plt.figure(figsize = (15,15))
ax1 = fig.add_subplot(121)
ax1.set_xticks([])
ax1.set_yticks([])

ax1.set_title('Original Image')
ax1.imshow(image)

ax2 = fig.add_subplot(122)
ax2.set_xticks([])
ax2.set_yticks([])

ax2.set_title('Canny Edges')
ax2.imshow(edges, cmap='gray')
Out[13]:
<matplotlib.image.AxesImage at 0x7ff64b8fa0b8>

Without first blurring the image, and removing small, local structures, a lot of irrelevant edge content gets picked up and amplified by the detector (as shown in the right panel above).

(IMPLEMENTATION) Blur the image then perform edge detection

In the next cell, you will repeat this experiment - blurring the image first to remove these local structures, so that only the important boudnary details remain in the edge-detected image.

Blur the image by using OpenCV's filter2d functionality - which is discussed in this documentation page - and use an averaging kernel of width equal to 4.

In [14]:
### TODO: Blur the test imageusing OpenCV's filter2d functionality, 
# Use an averaging kernel, and a kernel width equal to 4
kernel = np.ones((4,4),np.float32)/16
image_blured = cv2.filter2D(image,-1,kernel)

## TODO: Then perform Canny edge detection and display the output
# Convert to grayscale
gray_blured = cv2.cvtColor(image_blured, cv2.COLOR_RGB2GRAY)  

# Perform Canny edge detection
edges_blured = cv2.Canny(gray_blured,100,200)

# Dilate the image to amplify edges
edges_blured = cv2.dilate(edges_blured, None)

# Plot the RGB and edge-detected image
fig = plt.figure(figsize = (15,15))
ax1 = fig.add_subplot(121)
ax1.set_xticks([])
ax1.set_yticks([])

ax1.set_title('Original Image')
ax1.imshow(image)

ax2 = fig.add_subplot(122)
ax2.set_xticks([])
ax2.set_yticks([])

ax2.set_title('Canny Edges')
ax2.imshow(edges_blured, cmap='gray')
Out[14]:
<matplotlib.image.AxesImage at 0x7ff64b834668>

Step 4: Automatically Hide the Identity of an Individual

If you film something like a documentary or reality TV, you must get permission from every individual shown on film before you can show their face, otherwise you need to blur it out - by blurring the face a lot (so much so that even the global structures are obscured)! This is also true for projects like Google's StreetView maps - an enormous collection of mapping images taken from a fleet of Google vehicles. Because it would be impossible for Google to get the permission of every single person accidentally captured in one of these images they blur out everyone's faces, the detected images must automatically blur the identity of detected people. Here's a few examples of folks caught in the camera of a Google street view vehicle.

<img src="images/streetview_example_1.jpg" width=400 height=300/> <img src="images/streetview_example_2.jpg" width=400 height=300/>

Read in an image to perform identity detection

Let's try this out for ourselves. Use the face detection pipeline built above and what you know about using the filter2D to blur and image, and use these in tandem to hide the identity of the person in the following image - loaded in and printed in the next cell.

In [15]:
# Load in the image
image = cv2.imread('images/gus.jpg')

# Convert the image to RGB colorspace
image = cv2.cvtColor(image, cv2.COLOR_BGR2RGB)

# Display the image
fig = plt.figure(figsize = (6,6))
ax1 = fig.add_subplot(111)
ax1.set_xticks([])
ax1.set_yticks([])
ax1.set_title('Original Image')
ax1.imshow(image)
Out[15]:
<matplotlib.image.AxesImage at 0x7ff64b7babe0>

(IMPLEMENTATION) Use blurring to hide the identity of an individual in an image

The idea here is to 1) automatically detect the face in this image, and then 2) blur it out! Make sure to adjust the parameters of the averaging blur filter to completely obscure this person's identity.

In [16]:
## TODO: Implement face detection
# Convert the RGB  image to grayscale
gray = cv2.cvtColor(image, cv2.COLOR_RGB2GRAY)

# Extract the pre-trained face detector from an xml file
face_cascade = cv2.CascadeClassifier('detector_architectures/haarcascade_frontalface_default.xml')

# Detect the faces in image
faces = face_cascade.detectMultiScale(gray, 1.5, 4)

# Print the number of faces detected in the image
print('Number of faces detected:', len(faces))
   
## TODO: Blur the bounding box around each detected face using an averaging filter and display the result
# Make a copy of the orginal image to blur the bounding box areas
image_with_detections_blur = np.copy(image)

# Get the bounding box for each detected face
for (x,y,w,h) in faces:
    # averaging kernel with width equal to 20% of the max dimention of the bounding box
    k_width = int(max(w,h)/5)
    kernel = np.ones((k_width,k_width),np.float32)/(k_width*k_width)
    # Blur bounding box
    image_with_detections_blur[y:y+h, x:x+w] = cv2.filter2D(image_with_detections_blur[y:y+h, x:x+w],-1,kernel)
    #cv2.rectangle(image_with_detections_blur, (x,y), (x+w,y+h), (255,0,0), 3)
    
# Plot the original RGB and face blured images
fig = plt.figure(figsize = (15,15))
ax1 = fig.add_subplot(121)
ax1.set_xticks([])
ax1.set_yticks([])

ax1.set_title('Original Image')
ax1.imshow(image)

ax2 = fig.add_subplot(122)
ax2.set_xticks([])
ax2.set_yticks([])

ax2.set_title('Blured Identity Image')
ax2.imshow(image_with_detections_blur)
Number of faces detected: 1
Out[16]:
<matplotlib.image.AxesImage at 0x7ff64b6960f0>

(Optional) Build identity protection into your laptop camera

In this optional task you can add identity protection to your laptop camera, using the previously completed code where you added face detection to your laptop camera - and the task above. You should be able to get reasonable results with little parameter tuning - like the one shown in the gif below.

<img src="images/laptop_blurer_example.gif" width=400 height=300/>

As with the previous video task, to make this perfect would require significant effort - so don't strive for perfection here, strive for reasonable quality.

The next cell contains code a wrapper function called laptop_camera_identity_hider that - when called - will activate your laptop's camera. You need to place the relevant face detection and blurring code developed above in this function in order to blur faces entering your laptop camera's field of view.

Before adding anything to the function you can call it to get a hang of how it works - a small window will pop up showing you the live feed from your camera, you can press any key to close this window.

Note: Mac users may find that activating this function kills the kernel of their notebook every once in a while. If this happens to you, just restart your notebook's kernel, activate cell(s) containing any crucial import statements, and you'll be good to go!

In [17]:
### Insert face detection and blurring code into the wrapper below to create an identity protector on your laptop!
import cv2
import time 

def laptop_camera_go():
    # Create instance of video capturer
    cv2.namedWindow("face detection activated")
    vc = cv2.VideoCapture(0)

    # Try to get the first frame
    if vc.isOpened(): 
        rval, frame = vc.read()
    else:
        rval = False
    
    # Keep video stream open
    while rval:
        
        
        # Convert the RGB  image to grayscale
        gray = cv2.cvtColor(frame, cv2.COLOR_RGB2GRAY)
        
        # Detect the faces in image
        faces = face_cascade.detectMultiScale(gray, 1.25, 5)
        
        # Make a copy of the orginal image to draw face detections on
        frame_with_detections = np.copy(frame)
        
        ## Iterate faces and draw bounding box, detect eyes and draw bounding boxes for eyes in green
        eyes = []
        for (x,y,w,h) in faces:
            # averaging kernel with width equal to 20% of the max dimention of the bounding box
            k_width = int(max(w,h)/5)
            kernel = np.ones((k_width,k_width),np.float32)/(k_width*k_width)
            # Blur bounding box
            frame_with_detections[y:y+h, x:x+w] = cv2.filter2D(frame_with_detections[y:y+h, x:x+w],-1,kernel)
        
        # Plot the image from camera with identity protection
        cv2.imshow("face detection activated", frame_with_detections)
        
        # Exit functionality - press any key to exit laptop video
        key = cv2.waitKey(2)
        if key < 255: # Exit by pressing any key
            # Destroy windows
            cv2.destroyAllWindows()
            
            for i in range (1,5):
                cv2.waitKey(1)
            return
        
        # Read next frame
        time.sleep(0.05)             # control framerate for computation - default 20 frames per sec
        rval, frame = vc.read()    
        
In [19]:
# Run laptop identity hider
laptop_camera_go()

Step 5: Create a CNN to Recognize Facial Keypoints

OpenCV is often used in practice with other machine learning and deep learning libraries to produce interesting results. In this stage of the project you will create your own end-to-end pipeline - employing convolutional networks in keras along with OpenCV - to apply a "selfie" filter to streaming video and images.

You will start by creating and then training a convolutional network that can detect facial keypoints in a small dataset of cropped images of human faces. We then guide you towards OpenCV to expanding your detection algorithm to more general images. What are facial keypoints? Let's take a look at some examples.

<img src="images/keypoints_test_results.png" width=400 height=300/>

Facial keypoints (also called facial landmarks) are the small blue-green dots shown on each of the faces in the image above - there are 15 keypoints marked in each image. They mark important areas of the face - the eyes, corners of the mouth, the nose, etc. Facial keypoints can be used in a variety of machine learning applications from face and emotion recognition to commercial applications like the image filters popularized by Snapchat.

Below we illustrate a filter that, using the results of this section, automatically places sunglasses on people in images (using the facial keypoints to place the glasses correctly on each face). Here, the facial keypoints have been colored lime green for visualization purposes.

<img src="images/obamas_with_shades.png" width=1000 height=1000/>

Make a facial keypoint detector

But first things first: how can we make a facial keypoint detector? Well, at a high level, notice that facial keypoint detection is a regression problem. A single face corresponds to a set of 15 facial keypoints (a set of 15 corresponding $(x, y)$ coordinates, i.e., an output point). Because our input data are images, we can employ a convolutional neural network to recognize patterns in our images and learn how to identify these keypoint given sets of labeled data.

In order to train a regressor, we need a training set - a set of facial image / facial keypoint pairs to train on. For this we will be using this dataset from Kaggle. We've already downloaded this data and placed it in the data directory. Make sure that you have both the training and test data files. The training dataset contains several thousand $96 \times 96$ grayscale images of cropped human faces, along with each face's 15 corresponding facial keypoints (also called landmarks) that have been placed by hand, and recorded in $(x, y)$ coordinates. This wonderful resource also has a substantial testing set, which we will use in tinkering with our convolutional network.

To load in this data, run the Python cell below - notice we will load in both the training and testing sets.

The load_data function is in the included utils.py file.

In [18]:
# Load tensorflow
import tensorflow as tf
# set memory allocation base on demand from tensorflow and keras model
from keras import backend as K
config = tf.ConfigProto()
config.gpu_options.allow_growth=True
sess = tf.Session(config=config)
K.set_session(sess)
Using TensorFlow backend.
In [4]:
from utils import *

# Load training set
X_train, y_train = load_data()
print("X_train.shape == {}".format(X_train.shape))
print("y_train.shape == {}; y_train.min == {:.3f}; y_train.max == {:.3f}".format(
    y_train.shape, y_train.min(), y_train.max()))

# Load testing set
X_test, _ = load_data(test=True)
print("X_test.shape == {}".format(X_test.shape))
X_train.shape == (2140, 96, 96, 1)
y_train.shape == (2140, 30); y_train.min == -0.920; y_train.max == 0.996
X_test.shape == (1783, 96, 96, 1)

The load_data function in utils.py originates from this excellent blog post, which you are strongly encouraged to read. Please take the time now to review this function. Note how the output values - that is, the coordinates of each set of facial landmarks - have been normalized to take on values in the range $[-1, 1]$, while the pixel values of each input point (a facial image) have been normalized to the range $[0,1]$.

Note: the original Kaggle dataset contains some images with several missing keypoints. For simplicity, the load_data function removes those images with missing labels from the dataset. As an optional extension, you are welcome to amend the load_data function to include the incomplete data points.

Visualize the Training Data

Execute the code cell below to visualize a subset of the training data.

In [20]:
import matplotlib.pyplot as plt
%matplotlib inline

fig = plt.figure(figsize=(20,20))
fig.subplots_adjust(left=0, right=1, bottom=0, top=1, hspace=0.05, wspace=0.05)
for i in range(9):
    ax = fig.add_subplot(3, 3, i + 1, xticks=[], yticks=[])
    plot_data(X_train[i], y_train[i], ax)

For each training image, there are two landmarks per eyebrow (four total), three per eye (six total), four for the mouth, and one for the tip of the nose.

Review the plot_data function in utils.py to understand how the 30-dimensional training labels in y_train are mapped to facial locations, as this function will prove useful for your pipeline.

(IMPLEMENTATION) Specify the CNN Architecture

In this section, you will specify a neural network for predicting the locations of facial keypoints. Use the code cell below to specify the architecture of your neural network. We have imported some layers that you may find useful for this task, but if you need to use more Keras layers, feel free to import them in the cell.

Your network should accept a $96 \times 96$ grayscale image as input, and it should output a vector with 30 entries, corresponding to the predicted (horizontal and vertical) locations of 15 facial keypoints. If you are not sure where to start, you can find some useful starting architectures in this blog, but you are not permitted to copy any of the architectures that you find online.

In [94]:
# Import deep learning resources from Keras
from keras.models import Sequential
from keras.layers import Conv2D, MaxPooling2D, Dropout
from keras.layers import Flatten, Dense
from keras.layers.normalization import BatchNormalization


## TODO: Specify a CNN architecture
# Your model should accept 96x96 pixel graysale images in
# It should have a fully-connected output layer with 30 values (2 for each facial keypoint)

model = Sequential()

model.add(Conv2D(filters=16, kernel_size=3, padding='same', activation='relu', 
                        input_shape=(96, 96, 1)))
model.add(BatchNormalization())
model.add(Conv2D(filters=16, kernel_size=3, padding='same', activation='relu'))
model.add(BatchNormalization())
model.add(MaxPooling2D(pool_size=3))
model.add(Conv2D(filters=32, kernel_size=3, padding='same', activation='relu'))
model.add(BatchNormalization())
model.add(Conv2D(filters=32, kernel_size=3, padding='same', activation='relu'))
model.add(BatchNormalization())
model.add(MaxPooling2D(pool_size=3))
model.add(Conv2D(filters=64, kernel_size=3, padding='same', activation='relu'))
model.add(BatchNormalization())
model.add(Conv2D(filters=64, kernel_size=3, padding='same', activation='relu'))
model.add(BatchNormalization())
model.add(MaxPooling2D(pool_size=3))
model.add(Conv2D(filters=128, kernel_size=3, padding='same', activation='relu'))
model.add(BatchNormalization())
model.add(Conv2D(filters=128, kernel_size=3, padding='same', activation='relu'))
model.add(BatchNormalization())
model.add(MaxPooling2D(pool_size=3))
model.add(Flatten())
model.add(Dropout(0.25))
model.add(Dense(512, activation='relu'))
model.add(Dropout(0.25))
model.add(Dense(30))


# Summarize the model
model.summary()
_________________________________________________________________
Layer (type)                 Output Shape              Param #   
=================================================================
conv2d_187 (Conv2D)          (None, 96, 96, 16)        160       
_________________________________________________________________
batch_normalization_152 (Bat (None, 96, 96, 16)        64        
_________________________________________________________________
conv2d_188 (Conv2D)          (None, 96, 96, 16)        2320      
_________________________________________________________________
batch_normalization_153 (Bat (None, 96, 96, 16)        64        
_________________________________________________________________
max_pooling2d_108 (MaxPoolin (None, 32, 32, 16)        0         
_________________________________________________________________
conv2d_189 (Conv2D)          (None, 32, 32, 32)        4640      
_________________________________________________________________
batch_normalization_154 (Bat (None, 32, 32, 32)        128       
_________________________________________________________________
conv2d_190 (Conv2D)          (None, 32, 32, 32)        9248      
_________________________________________________________________
batch_normalization_155 (Bat (None, 32, 32, 32)        128       
_________________________________________________________________
max_pooling2d_109 (MaxPoolin (None, 10, 10, 32)        0         
_________________________________________________________________
conv2d_191 (Conv2D)          (None, 10, 10, 64)        18496     
_________________________________________________________________
batch_normalization_156 (Bat (None, 10, 10, 64)        256       
_________________________________________________________________
conv2d_192 (Conv2D)          (None, 10, 10, 64)        36928     
_________________________________________________________________
batch_normalization_157 (Bat (None, 10, 10, 64)        256       
_________________________________________________________________
max_pooling2d_110 (MaxPoolin (None, 3, 3, 64)          0         
_________________________________________________________________
conv2d_193 (Conv2D)          (None, 3, 3, 128)         73856     
_________________________________________________________________
batch_normalization_158 (Bat (None, 3, 3, 128)         512       
_________________________________________________________________
conv2d_194 (Conv2D)          (None, 3, 3, 128)         147584    
_________________________________________________________________
batch_normalization_159 (Bat (None, 3, 3, 128)         512       
_________________________________________________________________
max_pooling2d_111 (MaxPoolin (None, 1, 1, 128)         0         
_________________________________________________________________
flatten_26 (Flatten)         (None, 128)               0         
_________________________________________________________________
dropout_51 (Dropout)         (None, 128)               0         
_________________________________________________________________
dense_51 (Dense)             (None, 512)               66048     
_________________________________________________________________
dropout_52 (Dropout)         (None, 512)               0         
_________________________________________________________________
dense_52 (Dense)             (None, 30)                15390     
=================================================================
Total params: 376,590
Trainable params: 375,630
Non-trainable params: 960
_________________________________________________________________

Step 6: Compile and Train the Model

After specifying your architecture, you'll need to compile and train the model to detect facial keypoints'

(IMPLEMENTATION) Compile and Train the Model

Use the compile method to configure the learning process. Experiment with your choice of optimizer; you may have some ideas about which will work best (SGD vs. RMSprop, etc), but take the time to empirically verify your theories.

Use the fit method to train the model. Break off a validation set by setting validation_split=0.2. Save the returned History object in the history variable.

Your model is required to attain a validation loss (measured as mean squared error) of at least XYZ. When you have finished training, save your model as an HDF5 file with file path my_model.h5.

In [96]:
from keras.optimizers import SGD, RMSprop, Adagrad, Adadelta, Adam, Adamax, Nadam
from keras.callbacks import ModelCheckpoint, EarlyStopping

## TODO: Compile the model
model.compile(loss='mean_squared_error', optimizer='SGD')

## TODO: Train the model
epochs = 20
batch_size = 107

earlystopping = EarlyStopping(monitor='val_loss', min_delta=0, patience=10, verbose=0, mode='auto')
## TODO: Save the model as model.h5
# Save the model for best validation loss
checkpointer = ModelCheckpoint(filepath='my_model.h5', 
                               verbose=1, save_best_only=True)

hist_SGD = model.fit(X_train, y_train, validation_split=0.2, 
                 epochs=epochs, batch_size=batch_size, 
                 callbacks=[checkpointer, earlystopping], shuffle=True, verbose=1)
Train on 1712 samples, validate on 428 samples
Epoch 1/20
1605/1712 [===========================>..] - ETA: 1s - loss: 0.5638Epoch 00000: val_loss improved from inf to 0.14574, saving model to my_model.h5
1712/1712 [==============================] - 22s - loss: 0.5587 - val_loss: 0.1457
Epoch 2/20
1605/1712 [===========================>..] - ETA: 1s - loss: 0.4568Epoch 00001: val_loss improved from 0.14574 to 0.14066, saving model to my_model.h5
1712/1712 [==============================] - 19s - loss: 0.4539 - val_loss: 0.1407
Epoch 3/20
1605/1712 [===========================>..] - ETA: 1s - loss: 0.3735Epoch 00002: val_loss improved from 0.14066 to 0.13709, saving model to my_model.h5
1712/1712 [==============================] - 19s - loss: 0.3726 - val_loss: 0.1371
Epoch 4/20
1605/1712 [===========================>..] - ETA: 1s - loss: 0.3204Epoch 00003: val_loss improved from 0.13709 to 0.13403, saving model to my_model.h5
1712/1712 [==============================] - 19s - loss: 0.3187 - val_loss: 0.1340
Epoch 5/20
1605/1712 [===========================>..] - ETA: 1s - loss: 0.2731Epoch 00004: val_loss improved from 0.13403 to 0.13061, saving model to my_model.h5
1712/1712 [==============================] - 19s - loss: 0.2709 - val_loss: 0.1306
Epoch 6/20
1605/1712 [===========================>..] - ETA: 1s - loss: 0.2463Epoch 00005: val_loss improved from 0.13061 to 0.12786, saving model to my_model.h5
1712/1712 [==============================] - 19s - loss: 0.2461 - val_loss: 0.1279
Epoch 7/20
1605/1712 [===========================>..] - ETA: 1s - loss: 0.2193Epoch 00006: val_loss improved from 0.12786 to 0.12652, saving model to my_model.h5
1712/1712 [==============================] - 19s - loss: 0.2184 - val_loss: 0.1265
Epoch 8/20
1605/1712 [===========================>..] - ETA: 1s - loss: 0.1950Epoch 00007: val_loss improved from 0.12652 to 0.12623, saving model to my_model.h5
1712/1712 [==============================] - 19s - loss: 0.1943 - val_loss: 0.1262
Epoch 9/20
1605/1712 [===========================>..] - ETA: 1s - loss: 0.1772Epoch 00008: val_loss improved from 0.12623 to 0.12491, saving model to my_model.h5
1712/1712 [==============================] - 19s - loss: 0.1758 - val_loss: 0.1249
Epoch 10/20
1605/1712 [===========================>..] - ETA: 1s - loss: 0.1598Epoch 00009: val_loss improved from 0.12491 to 0.12247, saving model to my_model.h5
1712/1712 [==============================] - 19s - loss: 0.1605 - val_loss: 0.1225
Epoch 11/20
1605/1712 [===========================>..] - ETA: 1s - loss: 0.1484Epoch 00010: val_loss improved from 0.12247 to 0.11950, saving model to my_model.h5
1712/1712 [==============================] - 19s - loss: 0.1476 - val_loss: 0.1195
Epoch 12/20
1605/1712 [===========================>..] - ETA: 1s - loss: 0.1383Epoch 00011: val_loss improved from 0.11950 to 0.11425, saving model to my_model.h5
1712/1712 [==============================] - 19s - loss: 0.1379 - val_loss: 0.1142
Epoch 13/20
1605/1712 [===========================>..] - ETA: 1s - loss: 0.1300Epoch 00012: val_loss improved from 0.11425 to 0.10790, saving model to my_model.h5
1712/1712 [==============================] - 19s - loss: 0.1299 - val_loss: 0.1079
Epoch 14/20
1605/1712 [===========================>..] - ETA: 1s - loss: 0.1203Epoch 00013: val_loss improved from 0.10790 to 0.10244, saving model to my_model.h5
1712/1712 [==============================] - 19s - loss: 0.1202 - val_loss: 0.1024
Epoch 15/20
1605/1712 [===========================>..] - ETA: 1s - loss: 0.1137Epoch 00014: val_loss improved from 0.10244 to 0.09585, saving model to my_model.h5
1712/1712 [==============================] - 19s - loss: 0.1128 - val_loss: 0.0958
Epoch 16/20
1605/1712 [===========================>..] - ETA: 1s - loss: 0.1080Epoch 00015: val_loss improved from 0.09585 to 0.08900, saving model to my_model.h5
1712/1712 [==============================] - 19s - loss: 0.1079 - val_loss: 0.0890
Epoch 17/20
1605/1712 [===========================>..] - ETA: 1s - loss: 0.1036Epoch 00016: val_loss improved from 0.08900 to 0.08154, saving model to my_model.h5
1712/1712 [==============================] - 19s - loss: 0.1027 - val_loss: 0.0815
Epoch 18/20
1605/1712 [===========================>..] - ETA: 1s - loss: 0.0970Epoch 00017: val_loss improved from 0.08154 to 0.07602, saving model to my_model.h5
1712/1712 [==============================] - 19s - loss: 0.0970 - val_loss: 0.0760
Epoch 19/20
1605/1712 [===========================>..] - ETA: 1s - loss: 0.0927Epoch 00018: val_loss improved from 0.07602 to 0.07023, saving model to my_model.h5
1712/1712 [==============================] - 19s - loss: 0.0924 - val_loss: 0.0702
Epoch 20/20
1605/1712 [===========================>..] - ETA: 1s - loss: 0.0869Epoch 00019: val_loss improved from 0.07023 to 0.06288, saving model to my_model.h5
1712/1712 [==============================] - 19s - loss: 0.0871 - val_loss: 0.0629
In [109]:
#function to plot training history
def plot_train_history(model):
    """
    input: model training history from keras
    output: plot loss vs epochs
    """
    lw = 2
    plt.figure(figsize=(8, 6))
    plt.plot(model.epoch, model.history['loss'],c='b', label="Train loss", lw=lw)
    plt.plot(model.epoch, model.history['val_loss'],c='g', label="Validation loss", lw=lw)
    #plt.ylim([0.0, max(model.history['loss'])])
    #plt.xlim([0, model.epoch[-1]])
    plt.xlabel('Epochs')
    plt.ylabel('Loss')
    plt.legend()
    plt.show()
In [110]:
plot_train_history(hist_SGD)
In [111]:
# Final training

## TODO: Compile the model
model.compile(loss='mean_squared_error', optimizer='Nadam')

## TODO: Train the model
epochs = 200
batch_size = 107

earlystopping = EarlyStopping(monitor='val_loss', min_delta=0, patience=10, verbose=0, mode='auto')
## TODO: Save the model as model.h5
# Save the model for best validation loss
checkpointer = ModelCheckpoint(filepath='my_model.h5', 
                               verbose=1, save_best_only=True)

hist = model.fit(X_train, y_train, validation_split=0.2, 
                 epochs=epochs, batch_size=batch_size, 
                 callbacks=[checkpointer, earlystopping], shuffle=True, verbose=1)
Train on 1712 samples, validate on 428 samples
Epoch 1/200
1605/1712 [===========================>..] - ETA: 1s - loss: 0.0704Epoch 00000: val_loss improved from inf to 0.01319, saving model to my_model.h5
1712/1712 [==============================] - 23s - loss: 0.0681 - val_loss: 0.0132
Epoch 2/200
1605/1712 [===========================>..] - ETA: 1s - loss: 0.0277Epoch 00001: val_loss improved from 0.01319 to 0.00788, saving model to my_model.h5
1712/1712 [==============================] - 19s - loss: 0.0273 - val_loss: 0.0079
Epoch 3/200
1605/1712 [===========================>..] - ETA: 1s - loss: 0.0189Epoch 00002: val_loss did not improve
1712/1712 [==============================] - 19s - loss: 0.0187 - val_loss: 0.0129
Epoch 4/200
1605/1712 [===========================>..] - ETA: 1s - loss: 0.0131Epoch 00003: val_loss did not improve
1712/1712 [==============================] - 19s - loss: 0.0131 - val_loss: 0.0179
Epoch 5/200
1605/1712 [===========================>..] - ETA: 1s - loss: 0.0125Epoch 00004: val_loss did not improve
1712/1712 [==============================] - 19s - loss: 0.0125 - val_loss: 0.0147
Epoch 6/200
1605/1712 [===========================>..] - ETA: 1s - loss: 0.0112Epoch 00005: val_loss did not improve
1712/1712 [==============================] - 19s - loss: 0.0112 - val_loss: 0.0096
Epoch 7/200
1605/1712 [===========================>..] - ETA: 1s - loss: 0.0100Epoch 00006: val_loss improved from 0.00788 to 0.00472, saving model to my_model.h5
1712/1712 [==============================] - 19s - loss: 0.0100 - val_loss: 0.0047
Epoch 8/200
1605/1712 [===========================>..] - ETA: 1s - loss: 0.0088Epoch 00007: val_loss did not improve
1712/1712 [==============================] - 19s - loss: 0.0088 - val_loss: 0.0063
Epoch 9/200
1605/1712 [===========================>..] - ETA: 1s - loss: 0.0085Epoch 00008: val_loss did not improve
1712/1712 [==============================] - 19s - loss: 0.0085 - val_loss: 0.0064
Epoch 10/200
1605/1712 [===========================>..] - ETA: 1s - loss: 0.0074Epoch 00009: val_loss improved from 0.00472 to 0.00374, saving model to my_model.h5
1712/1712 [==============================] - 19s - loss: 0.0073 - val_loss: 0.0037
Epoch 11/200
1605/1712 [===========================>..] - ETA: 1s - loss: 0.0070Epoch 00010: val_loss improved from 0.00374 to 0.00364, saving model to my_model.h5
1712/1712 [==============================] - 19s - loss: 0.0069 - val_loss: 0.0036
Epoch 12/200
1605/1712 [===========================>..] - ETA: 1s - loss: 0.0070Epoch 00011: val_loss did not improve
1712/1712 [==============================] - 19s - loss: 0.0070 - val_loss: 0.0042
Epoch 13/200
1605/1712 [===========================>..] - ETA: 1s - loss: 0.0076Epoch 00012: val_loss improved from 0.00364 to 0.00355, saving model to my_model.h5
1712/1712 [==============================] - 19s - loss: 0.0075 - val_loss: 0.0036
Epoch 14/200
1605/1712 [===========================>..] - ETA: 1s - loss: 0.0065Epoch 00013: val_loss improved from 0.00355 to 0.00324, saving model to my_model.h5
1712/1712 [==============================] - 19s - loss: 0.0064 - val_loss: 0.0032
Epoch 15/200
1605/1712 [===========================>..] - ETA: 1s - loss: 0.0063Epoch 00014: val_loss improved from 0.00324 to 0.00315, saving model to my_model.h5
1712/1712 [==============================] - 19s - loss: 0.0063 - val_loss: 0.0032
Epoch 16/200
1605/1712 [===========================>..] - ETA: 1s - loss: 0.0059Epoch 00015: val_loss improved from 0.00315 to 0.00305, saving model to my_model.h5
1712/1712 [==============================] - 19s - loss: 0.0059 - val_loss: 0.0030
Epoch 17/200
1605/1712 [===========================>..] - ETA: 1s - loss: 0.0060Epoch 00016: val_loss did not improve
1712/1712 [==============================] - 19s - loss: 0.0061 - val_loss: 0.0033
Epoch 18/200
1605/1712 [===========================>..] - ETA: 1s - loss: 0.0057Epoch 00017: val_loss did not improve
1712/1712 [==============================] - 19s - loss: 0.0057 - val_loss: 0.0032
Epoch 19/200
1605/1712 [===========================>..] - ETA: 1s - loss: 0.0056Epoch 00018: val_loss did not improve
1712/1712 [==============================] - 19s - loss: 0.0056 - val_loss: 0.0030
Epoch 20/200
1605/1712 [===========================>..] - ETA: 1s - loss: 0.0052Epoch 00019: val_loss did not improve
1712/1712 [==============================] - 19s - loss: 0.0052 - val_loss: 0.0040
Epoch 21/200
1605/1712 [===========================>..] - ETA: 1s - loss: 0.0049Epoch 00020: val_loss did not improve
1712/1712 [==============================] - 19s - loss: 0.0049 - val_loss: 0.0037
Epoch 22/200
1605/1712 [===========================>..] - ETA: 1s - loss: 0.0051Epoch 00021: val_loss did not improve
1712/1712 [==============================] - 19s - loss: 0.0051 - val_loss: 0.0040
Epoch 23/200
1605/1712 [===========================>..] - ETA: 1s - loss: 0.0050Epoch 00022: val_loss did not improve
1712/1712 [==============================] - 19s - loss: 0.0050 - val_loss: 0.0031
Epoch 24/200
1605/1712 [===========================>..] - ETA: 1s - loss: 0.0048Epoch 00023: val_loss did not improve
1712/1712 [==============================] - 19s - loss: 0.0048 - val_loss: 0.0041
Epoch 25/200
1605/1712 [===========================>..] - ETA: 1s - loss: 0.0049Epoch 00024: val_loss improved from 0.00305 to 0.00296, saving model to my_model.h5
1712/1712 [==============================] - 19s - loss: 0.0049 - val_loss: 0.0030
Epoch 26/200
1605/1712 [===========================>..] - ETA: 1s - loss: 0.0046Epoch 00025: val_loss improved from 0.00296 to 0.00272, saving model to my_model.h5
1712/1712 [==============================] - 19s - loss: 0.0046 - val_loss: 0.0027
Epoch 27/200
1605/1712 [===========================>..] - ETA: 1s - loss: 0.0046Epoch 00026: val_loss improved from 0.00272 to 0.00263, saving model to my_model.h5
1712/1712 [==============================] - 19s - loss: 0.0047 - val_loss: 0.0026
Epoch 28/200
1605/1712 [===========================>..] - ETA: 1s - loss: 0.0043Epoch 00027: val_loss did not improve
1712/1712 [==============================] - 19s - loss: 0.0043 - val_loss: 0.0036
Epoch 29/200
1605/1712 [===========================>..] - ETA: 1s - loss: 0.0041Epoch 00028: val_loss did not improve
1712/1712 [==============================] - 19s - loss: 0.0041 - val_loss: 0.0031
Epoch 30/200
1605/1712 [===========================>..] - ETA: 1s - loss: 0.0042Epoch 00029: val_loss did not improve
1712/1712 [==============================] - 19s - loss: 0.0042 - val_loss: 0.0027
Epoch 31/200
1605/1712 [===========================>..] - ETA: 1s - loss: 0.0042Epoch 00030: val_loss did not improve
1712/1712 [==============================] - 19s - loss: 0.0042 - val_loss: 0.0028
Epoch 32/200
1605/1712 [===========================>..] - ETA: 1s - loss: 0.0040Epoch 00031: val_loss improved from 0.00263 to 0.00257, saving model to my_model.h5
1712/1712 [==============================] - 19s - loss: 0.0039 - val_loss: 0.0026
Epoch 33/200
1605/1712 [===========================>..] - ETA: 1s - loss: 0.0040Epoch 00032: val_loss improved from 0.00257 to 0.00243, saving model to my_model.h5
1712/1712 [==============================] - 19s - loss: 0.0040 - val_loss: 0.0024
Epoch 34/200
1605/1712 [===========================>..] - ETA: 1s - loss: 0.0038Epoch 00033: val_loss improved from 0.00243 to 0.00239, saving model to my_model.h5
1712/1712 [==============================] - 19s - loss: 0.0038 - val_loss: 0.0024
Epoch 35/200
1605/1712 [===========================>..] - ETA: 1s - loss: 0.0039Epoch 00034: val_loss did not improve
1712/1712 [==============================] - 19s - loss: 0.0038 - val_loss: 0.0024
Epoch 36/200
1605/1712 [===========================>..] - ETA: 1s - loss: 0.0035Epoch 00035: val_loss did not improve
1712/1712 [==============================] - 19s - loss: 0.0036 - val_loss: 0.0028
Epoch 37/200
1605/1712 [===========================>..] - ETA: 1s - loss: 0.0036Epoch 00036: val_loss did not improve
1712/1712 [==============================] - 19s - loss: 0.0036 - val_loss: 0.0024
Epoch 38/200
1605/1712 [===========================>..] - ETA: 1s - loss: 0.0035Epoch 00037: val_loss improved from 0.00239 to 0.00236, saving model to my_model.h5
1712/1712 [==============================] - 19s - loss: 0.0035 - val_loss: 0.0024
Epoch 39/200
1605/1712 [===========================>..] - ETA: 1s - loss: 0.0035Epoch 00038: val_loss did not improve
1712/1712 [==============================] - 19s - loss: 0.0035 - val_loss: 0.0024
Epoch 40/200
1605/1712 [===========================>..] - ETA: 1s - loss: 0.0034Epoch 00039: val_loss did not improve
1712/1712 [==============================] - 19s - loss: 0.0034 - val_loss: 0.0027
Epoch 41/200
1605/1712 [===========================>..] - ETA: 1s - loss: 0.0033Epoch 00040: val_loss did not improve
1712/1712 [==============================] - 19s - loss: 0.0033 - val_loss: 0.0025
Epoch 42/200
1605/1712 [===========================>..] - ETA: 1s - loss: 0.0033Epoch 00041: val_loss did not improve
1712/1712 [==============================] - 19s - loss: 0.0033 - val_loss: 0.0025
Epoch 43/200
1605/1712 [===========================>..] - ETA: 1s - loss: 0.0033Epoch 00042: val_loss improved from 0.00236 to 0.00232, saving model to my_model.h5
1712/1712 [==============================] - 19s - loss: 0.0033 - val_loss: 0.0023
Epoch 44/200
1605/1712 [===========================>..] - ETA: 1s - loss: 0.0033Epoch 00043: val_loss improved from 0.00232 to 0.00220, saving model to my_model.h5
1712/1712 [==============================] - 19s - loss: 0.0032 - val_loss: 0.0022
Epoch 45/200
1605/1712 [===========================>..] - ETA: 1s - loss: 0.0031Epoch 00044: val_loss did not improve
1712/1712 [==============================] - 19s - loss: 0.0031 - val_loss: 0.0022
Epoch 46/200
1605/1712 [===========================>..] - ETA: 1s - loss: 0.0031Epoch 00045: val_loss did not improve
1712/1712 [==============================] - 19s - loss: 0.0031 - val_loss: 0.0022
Epoch 47/200
1605/1712 [===========================>..] - ETA: 1s - loss: 0.0030Epoch 00046: val_loss did not improve
1712/1712 [==============================] - 19s - loss: 0.0030 - val_loss: 0.0023
Epoch 48/200
1605/1712 [===========================>..] - ETA: 1s - loss: 0.0029Epoch 00047: val_loss did not improve
1712/1712 [==============================] - 19s - loss: 0.0030 - val_loss: 0.0022
Epoch 49/200
1605/1712 [===========================>..] - ETA: 1s - loss: 0.0030Epoch 00048: val_loss did not improve
1712/1712 [==============================] - 19s - loss: 0.0030 - val_loss: 0.0022
Epoch 50/200
1605/1712 [===========================>..] - ETA: 1s - loss: 0.0030Epoch 00049: val_loss improved from 0.00220 to 0.00212, saving model to my_model.h5
1712/1712 [==============================] - 19s - loss: 0.0030 - val_loss: 0.0021
Epoch 51/200
1605/1712 [===========================>..] - ETA: 1s - loss: 0.0029Epoch 00050: val_loss improved from 0.00212 to 0.00210, saving model to my_model.h5
1712/1712 [==============================] - 19s - loss: 0.0029 - val_loss: 0.0021
Epoch 52/200
1605/1712 [===========================>..] - ETA: 1s - loss: 0.0029Epoch 00051: val_loss did not improve
1712/1712 [==============================] - 19s - loss: 0.0029 - val_loss: 0.0022
Epoch 53/200
1605/1712 [===========================>..] - ETA: 1s - loss: 0.0028Epoch 00052: val_loss improved from 0.00210 to 0.00203, saving model to my_model.h5
1712/1712 [==============================] - 19s - loss: 0.0028 - val_loss: 0.0020
Epoch 54/200
1605/1712 [===========================>..] - ETA: 1s - loss: 0.0028Epoch 00053: val_loss did not improve
1712/1712 [==============================] - 19s - loss: 0.0027 - val_loss: 0.0024
Epoch 55/200
1605/1712 [===========================>..] - ETA: 1s - loss: 0.0027Epoch 00054: val_loss did not improve
1712/1712 [==============================] - 19s - loss: 0.0027 - val_loss: 0.0021
Epoch 56/200
1605/1712 [===========================>..] - ETA: 1s - loss: 0.0026Epoch 00055: val_loss did not improve
1712/1712 [==============================] - 19s - loss: 0.0026 - val_loss: 0.0024
Epoch 57/200
1605/1712 [===========================>..] - ETA: 1s - loss: 0.0027Epoch 00056: val_loss did not improve
1712/1712 [==============================] - 19s - loss: 0.0027 - val_loss: 0.0022
Epoch 58/200
1605/1712 [===========================>..] - ETA: 1s - loss: 0.0026Epoch 00057: val_loss did not improve
1712/1712 [==============================] - 19s - loss: 0.0026 - val_loss: 0.0023
Epoch 59/200
1605/1712 [===========================>..] - ETA: 1s - loss: 0.0026Epoch 00058: val_loss did not improve
1712/1712 [==============================] - 19s - loss: 0.0026 - val_loss: 0.0021
Epoch 60/200
1605/1712 [===========================>..] - ETA: 1s - loss: 0.0025Epoch 00059: val_loss did not improve
1712/1712 [==============================] - 19s - loss: 0.0025 - val_loss: 0.0021
Epoch 61/200
1605/1712 [===========================>..] - ETA: 1s - loss: 0.0025Epoch 00060: val_loss did not improve
1712/1712 [==============================] - 19s - loss: 0.0025 - val_loss: 0.0021
Epoch 62/200
1605/1712 [===========================>..] - ETA: 1s - loss: 0.0025Epoch 00061: val_loss did not improve
1712/1712 [==============================] - 19s - loss: 0.0024 - val_loss: 0.0023
Epoch 63/200
1605/1712 [===========================>..] - ETA: 1s - loss: 0.0025Epoch 00062: val_loss did not improve
1712/1712 [==============================] - 19s - loss: 0.0025 - val_loss: 0.0021
Epoch 64/200
1605/1712 [===========================>..] - ETA: 1s - loss: 0.0024Epoch 00063: val_loss improved from 0.00203 to 0.00201, saving model to my_model.h5
1712/1712 [==============================] - 19s - loss: 0.0024 - val_loss: 0.0020
Epoch 65/200
1605/1712 [===========================>..] - ETA: 1s - loss: 0.0023Epoch 00064: val_loss did not improve
1712/1712 [==============================] - 19s - loss: 0.0023 - val_loss: 0.0020
Epoch 66/200
1605/1712 [===========================>..] - ETA: 1s - loss: 0.0023Epoch 00065: val_loss did not improve
1712/1712 [==============================] - 19s - loss: 0.0023 - val_loss: 0.0022
Epoch 67/200
1605/1712 [===========================>..] - ETA: 1s - loss: 0.0023Epoch 00066: val_loss did not improve
1712/1712 [==============================] - 19s - loss: 0.0023 - val_loss: 0.0022
Epoch 68/200
1605/1712 [===========================>..] - ETA: 1s - loss: 0.0023Epoch 00067: val_loss did not improve
1712/1712 [==============================] - 19s - loss: 0.0023 - val_loss: 0.0021
Epoch 69/200
1605/1712 [===========================>..] - ETA: 1s - loss: 0.0023Epoch 00068: val_loss improved from 0.00201 to 0.00200, saving model to my_model.h5
1712/1712 [==============================] - 19s - loss: 0.0023 - val_loss: 0.0020
Epoch 70/200
1605/1712 [===========================>..] - ETA: 1s - loss: 0.0023Epoch 00069: val_loss did not improve
1712/1712 [==============================] - 19s - loss: 0.0023 - val_loss: 0.0021
Epoch 71/200
1605/1712 [===========================>..] - ETA: 1s - loss: 0.0022Epoch 00070: val_loss improved from 0.00200 to 0.00194, saving model to my_model.h5
1712/1712 [==============================] - 19s - loss: 0.0022 - val_loss: 0.0019
Epoch 72/200
1605/1712 [===========================>..] - ETA: 1s - loss: 0.0022Epoch 00071: val_loss did not improve
1712/1712 [==============================] - 19s - loss: 0.0022 - val_loss: 0.0020
Epoch 73/200
1605/1712 [===========================>..] - ETA: 1s - loss: 0.0022Epoch 00072: val_loss did not improve
1712/1712 [==============================] - 19s - loss: 0.0022 - val_loss: 0.0020
Epoch 74/200
1605/1712 [===========================>..] - ETA: 1s - loss: 0.0022Epoch 00073: val_loss did not improve
1712/1712 [==============================] - 19s - loss: 0.0022 - val_loss: 0.0020
Epoch 75/200
1605/1712 [===========================>..] - ETA: 1s - loss: 0.0021Epoch 00074: val_loss did not improve
1712/1712 [==============================] - 19s - loss: 0.0021 - val_loss: 0.0020
Epoch 76/200
1605/1712 [===========================>..] - ETA: 1s - loss: 0.0021Epoch 00075: val_loss improved from 0.00194 to 0.00190, saving model to my_model.h5
1712/1712 [==============================] - 19s - loss: 0.0021 - val_loss: 0.0019
Epoch 77/200
1605/1712 [===========================>..] - ETA: 1s - loss: 0.0021Epoch 00076: val_loss did not improve
1712/1712 [==============================] - 19s - loss: 0.0021 - val_loss: 0.0020
Epoch 78/200
1605/1712 [===========================>..] - ETA: 1s - loss: 0.0021Epoch 00077: val_loss improved from 0.00190 to 0.00178, saving model to my_model.h5
1712/1712 [==============================] - 19s - loss: 0.0021 - val_loss: 0.0018
Epoch 79/200
1605/1712 [===========================>..] - ETA: 1s - loss: 0.0021Epoch 00078: val_loss did not improve
1712/1712 [==============================] - 19s - loss: 0.0021 - val_loss: 0.0018
Epoch 80/200
1605/1712 [===========================>..] - ETA: 1s - loss: 0.0020Epoch 00079: val_loss did not improve
1712/1712 [==============================] - 19s - loss: 0.0020 - val_loss: 0.0018
Epoch 81/200
1605/1712 [===========================>..] - ETA: 1s - loss: 0.0020Epoch 00080: val_loss did not improve
1712/1712 [==============================] - 19s - loss: 0.0020 - val_loss: 0.0019
Epoch 82/200
1605/1712 [===========================>..] - ETA: 1s - loss: 0.0020Epoch 00081: val_loss did not improve
1712/1712 [==============================] - 19s - loss: 0.0020 - val_loss: 0.0018
Epoch 83/200
1605/1712 [===========================>..] - ETA: 1s - loss: 0.0020Epoch 00082: val_loss improved from 0.00178 to 0.00168, saving model to my_model.h5
1712/1712 [==============================] - 19s - loss: 0.0020 - val_loss: 0.0017
Epoch 84/200
1605/1712 [===========================>..] - ETA: 1s - loss: 0.0020Epoch 00083: val_loss did not improve
1712/1712 [==============================] - 19s - loss: 0.0020 - val_loss: 0.0017
Epoch 85/200
1605/1712 [===========================>..] - ETA: 1s - loss: 0.0020Epoch 00084: val_loss did not improve
1712/1712 [==============================] - 19s - loss: 0.0020 - val_loss: 0.0018
Epoch 86/200
1605/1712 [===========================>..] - ETA: 1s - loss: 0.0020Epoch 00085: val_loss did not improve
1712/1712 [==============================] - 19s - loss: 0.0020 - val_loss: 0.0018
Epoch 87/200
1605/1712 [===========================>..] - ETA: 1s - loss: 0.0020Epoch 00086: val_loss improved from 0.00168 to 0.00164, saving model to my_model.h5
1712/1712 [==============================] - 19s - loss: 0.0020 - val_loss: 0.0016
Epoch 88/200
1605/1712 [===========================>..] - ETA: 1s - loss: 0.0020Epoch 00087: val_loss did not improve
1712/1712 [==============================] - 19s - loss: 0.0019 - val_loss: 0.0017
Epoch 89/200
1605/1712 [===========================>..] - ETA: 1s - loss: 0.0019Epoch 00088: val_loss did not improve
1712/1712 [==============================] - 19s - loss: 0.0019 - val_loss: 0.0017
Epoch 90/200
1605/1712 [===========================>..] - ETA: 1s - loss: 0.0019Epoch 00089: val_loss did not improve
1712/1712 [==============================] - 19s - loss: 0.0019 - val_loss: 0.0018
Epoch 91/200
1605/1712 [===========================>..] - ETA: 1s - loss: 0.0019Epoch 00090: val_loss did not improve
1712/1712 [==============================] - 19s - loss: 0.0019 - val_loss: 0.0017
Epoch 92/200
1605/1712 [===========================>..] - ETA: 1s - loss: 0.0019Epoch 00091: val_loss improved from 0.00164 to 0.00162, saving model to my_model.h5
1712/1712 [==============================] - 19s - loss: 0.0019 - val_loss: 0.0016
Epoch 93/200
1605/1712 [===========================>..] - ETA: 1s - loss: 0.0019Epoch 00092: val_loss improved from 0.00162 to 0.00162, saving model to my_model.h5
1712/1712 [==============================] - 19s - loss: 0.0019 - val_loss: 0.0016
Epoch 94/200
1605/1712 [===========================>..] - ETA: 1s - loss: 0.0019Epoch 00093: val_loss did not improve
1712/1712 [==============================] - 19s - loss: 0.0019 - val_loss: 0.0017
Epoch 95/200
1605/1712 [===========================>..] - ETA: 1s - loss: 0.0019Epoch 00094: val_loss improved from 0.00162 to 0.00161, saving model to my_model.h5
1712/1712 [==============================] - 19s - loss: 0.0019 - val_loss: 0.0016
Epoch 96/200
1605/1712 [===========================>..] - ETA: 1s - loss: 0.0018Epoch 00095: val_loss improved from 0.00161 to 0.00155, saving model to my_model.h5
1712/1712 [==============================] - 19s - loss: 0.0018 - val_loss: 0.0015
Epoch 97/200
1605/1712 [===========================>..] - ETA: 1s - loss: 0.0019Epoch 00096: val_loss did not improve
1712/1712 [==============================] - 19s - loss: 0.0019 - val_loss: 0.0016
Epoch 98/200
1605/1712 [===========================>..] - ETA: 1s - loss: 0.0018Epoch 00097: val_loss did not improve
1712/1712 [==============================] - 19s - loss: 0.0018 - val_loss: 0.0017
Epoch 99/200
1605/1712 [===========================>..] - ETA: 1s - loss: 0.0018Epoch 00098: val_loss did not improve
1712/1712 [==============================] - 19s - loss: 0.0018 - val_loss: 0.0017
Epoch 100/200
1605/1712 [===========================>..] - ETA: 1s - loss: 0.0018Epoch 00099: val_loss did not improve
1712/1712 [==============================] - 19s - loss: 0.0018 - val_loss: 0.0017
Epoch 101/200
1605/1712 [===========================>..] - ETA: 1s - loss: 0.0018Epoch 00100: val_loss did not improve
1712/1712 [==============================] - 19s - loss: 0.0018 - val_loss: 0.0016
Epoch 102/200
1605/1712 [===========================>..] - ETA: 1s - loss: 0.0018Epoch 00101: val_loss did not improve
1712/1712 [==============================] - 19s - loss: 0.0018 - val_loss: 0.0016
Epoch 103/200
1605/1712 [===========================>..] - ETA: 1s - loss: 0.0019Epoch 00102: val_loss did not improve
1712/1712 [==============================] - 19s - loss: 0.0019 - val_loss: 0.0017
Epoch 104/200
1605/1712 [===========================>..] - ETA: 1s - loss: 0.0018Epoch 00103: val_loss did not improve
1712/1712 [==============================] - 19s - loss: 0.0018 - val_loss: 0.0017
Epoch 105/200
1605/1712 [===========================>..] - ETA: 1s - loss: 0.0018Epoch 00104: val_loss did not improve
1712/1712 [==============================] - 19s - loss: 0.0018 - val_loss: 0.0016
Epoch 106/200
1605/1712 [===========================>..] - ETA: 1s - loss: 0.0017Epoch 00105: val_loss improved from 0.00155 to 0.00149, saving model to my_model.h5
1712/1712 [==============================] - 19s - loss: 0.0017 - val_loss: 0.0015
Epoch 107/200
1605/1712 [===========================>..] - ETA: 1s - loss: 0.0018Epoch 00106: val_loss did not improve
1712/1712 [==============================] - 19s - loss: 0.0018 - val_loss: 0.0015
Epoch 108/200
1605/1712 [===========================>..] - ETA: 1s - loss: 0.0017Epoch 00107: val_loss did not improve
1712/1712 [==============================] - 19s - loss: 0.0017 - val_loss: 0.0015
Epoch 109/200
1605/1712 [===========================>..] - ETA: 1s - loss: 0.0017Epoch 00108: val_loss did not improve
1712/1712 [==============================] - 19s - loss: 0.0017 - val_loss: 0.0016
Epoch 110/200
1605/1712 [===========================>..] - ETA: 1s - loss: 0.0017Epoch 00109: val_loss did not improve
1712/1712 [==============================] - 19s - loss: 0.0017 - val_loss: 0.0016
Epoch 111/200
1605/1712 [===========================>..] - ETA: 1s - loss: 0.0018Epoch 00110: val_loss did not improve
1712/1712 [==============================] - 19s - loss: 0.0018 - val_loss: 0.0016
Epoch 112/200
1605/1712 [===========================>..] - ETA: 1s - loss: 0.0017Epoch 00111: val_loss did not improve
1712/1712 [==============================] - 19s - loss: 0.0017 - val_loss: 0.0015
Epoch 113/200
1605/1712 [===========================>..] - ETA: 1s - loss: 0.0017Epoch 00112: val_loss did not improve
1712/1712 [==============================] - 19s - loss: 0.0017 - val_loss: 0.0015
Epoch 114/200
1605/1712 [===========================>..] - ETA: 1s - loss: 0.0017Epoch 00113: val_loss did not improve
1712/1712 [==============================] - 19s - loss: 0.0017 - val_loss: 0.0015
Epoch 115/200
1605/1712 [===========================>..] - ETA: 1s - loss: 0.0017Epoch 00114: val_loss did not improve
1712/1712 [==============================] - 19s - loss: 0.0018 - val_loss: 0.0015
Epoch 116/200
1605/1712 [===========================>..] - ETA: 1s - loss: 0.0017Epoch 00115: val_loss did not improve
1712/1712 [==============================] - 19s - loss: 0.0017 - val_loss: 0.0016
Epoch 117/200
1605/1712 [===========================>..] - ETA: 1s - loss: 0.0016Epoch 00116: val_loss improved from 0.00149 to 0.00144, saving model to my_model.h5
1712/1712 [==============================] - 19s - loss: 0.0017 - val_loss: 0.0014
Epoch 118/200
1605/1712 [===========================>..] - ETA: 1s - loss: 0.0017Epoch 00117: val_loss did not improve
1712/1712 [==============================] - 19s - loss: 0.0017 - val_loss: 0.0016
Epoch 119/200
1605/1712 [===========================>..] - ETA: 1s - loss: 0.0017Epoch 00118: val_loss did not improve
1712/1712 [==============================] - 19s - loss: 0.0017 - val_loss: 0.0015
Epoch 120/200
1605/1712 [===========================>..] - ETA: 1s - loss: 0.0017Epoch 00119: val_loss did not improve
1712/1712 [==============================] - 19s - loss: 0.0017 - val_loss: 0.0016
Epoch 121/200
1605/1712 [===========================>..] - ETA: 1s - loss: 0.0017Epoch 00120: val_loss did not improve
1712/1712 [==============================] - 19s - loss: 0.0017 - val_loss: 0.0015
Epoch 122/200
1605/1712 [===========================>..] - ETA: 1s - loss: 0.0016Epoch 00121: val_loss improved from 0.00144 to 0.00144, saving model to my_model.h5
1712/1712 [==============================] - 19s - loss: 0.0016 - val_loss: 0.0014
Epoch 123/200
1605/1712 [===========================>..] - ETA: 1s - loss: 0.0017Epoch 00122: val_loss did not improve
1712/1712 [==============================] - 19s - loss: 0.0016 - val_loss: 0.0015
Epoch 124/200
1605/1712 [===========================>..] - ETA: 1s - loss: 0.0017Epoch 00123: val_loss improved from 0.00144 to 0.00143, saving model to my_model.h5
1712/1712 [==============================] - 19s - loss: 0.0016 - val_loss: 0.0014
Epoch 125/200
1605/1712 [===========================>..] - ETA: 1s - loss: 0.0017Epoch 00124: val_loss did not improve
1712/1712 [==============================] - 19s - loss: 0.0017 - val_loss: 0.0015
Epoch 126/200
1605/1712 [===========================>..] - ETA: 1s - loss: 0.0016Epoch 00125: val_loss did not improve
1712/1712 [==============================] - 19s - loss: 0.0016 - val_loss: 0.0015
Epoch 127/200
1605/1712 [===========================>..] - ETA: 1s - loss: 0.0016Epoch 00126: val_loss did not improve
1712/1712 [==============================] - 19s - loss: 0.0016 - val_loss: 0.0015
Epoch 128/200
1605/1712 [===========================>..] - ETA: 1s - loss: 0.0016Epoch 00127: val_loss did not improve
1712/1712 [==============================] - 19s - loss: 0.0016 - val_loss: 0.0015
Epoch 129/200
1605/1712 [===========================>..] - ETA: 1s - loss: 0.0016Epoch 00128: val_loss did not improve
1712/1712 [==============================] - 19s - loss: 0.0016 - val_loss: 0.0015
Epoch 130/200
1605/1712 [===========================>..] - ETA: 1s - loss: 0.0017Epoch 00129: val_loss did not improve
1712/1712 [==============================] - 19s - loss: 0.0017 - val_loss: 0.0015
Epoch 131/200
1605/1712 [===========================>..] - ETA: 1s - loss: 0.0016Epoch 00130: val_loss did not improve
1712/1712 [==============================] - 19s - loss: 0.0016 - val_loss: 0.0015
Epoch 132/200
1605/1712 [===========================>..] - ETA: 1s - loss: 0.0017Epoch 00131: val_loss did not improve
1712/1712 [==============================] - 19s - loss: 0.0017 - val_loss: 0.0014
Epoch 133/200
1605/1712 [===========================>..] - ETA: 1s - loss: 0.0016Epoch 00132: val_loss improved from 0.00143 to 0.00137, saving model to my_model.h5
1712/1712 [==============================] - 19s - loss: 0.0016 - val_loss: 0.0014
Epoch 134/200
1605/1712 [===========================>..] - ETA: 1s - loss: 0.0017Epoch 00133: val_loss did not improve
1712/1712 [==============================] - 19s - loss: 0.0017 - val_loss: 0.0015
Epoch 135/200
1605/1712 [===========================>..] - ETA: 1s - loss: 0.0016Epoch 00134: val_loss did not improve
1712/1712 [==============================] - 19s - loss: 0.0016 - val_loss: 0.0014
Epoch 136/200
1605/1712 [===========================>..] - ETA: 1s - loss: 0.0016Epoch 00135: val_loss did not improve
1712/1712 [==============================] - 19s - loss: 0.0016 - val_loss: 0.0014
Epoch 137/200
1605/1712 [===========================>..] - ETA: 1s - loss: 0.0016Epoch 00136: val_loss did not improve
1712/1712 [==============================] - 19s - loss: 0.0016 - val_loss: 0.0014
Epoch 138/200
1605/1712 [===========================>..] - ETA: 1s - loss: 0.0016Epoch 00137: val_loss did not improve
1712/1712 [==============================] - 19s - loss: 0.0016 - val_loss: 0.0014
Epoch 139/200
1605/1712 [===========================>..] - ETA: 1s - loss: 0.0016Epoch 00138: val_loss did not improve
1712/1712 [==============================] - 19s - loss: 0.0016 - val_loss: 0.0014
Epoch 140/200
1605/1712 [===========================>..] - ETA: 1s - loss: 0.0016Epoch 00139: val_loss did not improve
1712/1712 [==============================] - 19s - loss: 0.0016 - val_loss: 0.0014
Epoch 141/200
1605/1712 [===========================>..] - ETA: 1s - loss: 0.0016Epoch 00140: val_loss did not improve
1712/1712 [==============================] - 19s - loss: 0.0016 - val_loss: 0.0014
Epoch 142/200
1605/1712 [===========================>..] - ETA: 1s - loss: 0.0016Epoch 00141: val_loss did not improve
1712/1712 [==============================] - 19s - loss: 0.0016 - val_loss: 0.0014
Epoch 143/200
1605/1712 [===========================>..] - ETA: 1s - loss: 0.0016Epoch 00142: val_loss did not improve
1712/1712 [==============================] - 19s - loss: 0.0016 - val_loss: 0.0014
Epoch 144/200
1605/1712 [===========================>..] - ETA: 1s - loss: 0.0016Epoch 00143: val_loss did not improve
1712/1712 [==============================] - 19s - loss: 0.0016 - val_loss: 0.0014

Step 7: Visualize the Loss and Test Predictions

(IMPLEMENTATION) Answer a few questions and visualize the loss

Question 1: Outline the steps you took to get to your final neural network architecture and your reasoning at each step.

Answer: The idea of my CNN’s architecture came from AlexNet and VGGNet architectures. Due to the time and computing power limitation (even I have trained my model using a NVIDIA GTX 1070 GPU) I tried to use the simplicity of shallow AlexNet with advances of VGGNet. Thus, instead of using one convolutional layer (big kernel) followed by a rectified unit and a spatial pooling, I have used two convolutional consecutive layers with smaller kernels (3x3, resulting in the bigger combined receptive field) followed by only one rectified unit and then a spatial Max pooling. A batch normalization layer were introduced after each convolutional layer to dynamically normalize the inputs on a per mini-batch basis, thus speeding up learning process with improved generalization (improving optimization with regularization effect). The last set of convolutional layers is followed by flatten layer. The convolutional layers were design to extract spacial information of pixels from the images in the manner that the last layer will bring in-plane resolution of images from 96x96 to multiple layers of 3x3 matrixes with depth increased from 1 to 128. Thus, the convolutional filters complexity were increased for each consecutive pair of layers (16, 32, 64, 128), thus the first coupled layers will extract the simplest features and the last pair the most complex features. The next layer is a dropout layer to avoid overfitting (regularization), thus improve generalization. On top of CNN, I have added a dense layer of 512 nodes to add non-linear function to extract more information after the spacial information extraction. One more dropout layer were added to improve model generalization. The last layer is a 30 nodes layer, because of this is a regression problem, thus the lost function used in the trading was a mean squared error (MSE).

Question 2: Defend your choice of optimizer. Which optimizers did you test, and how did you determine which worked best?

Answer: I have tested multiple optimizers and each one has it’s own prons and cons. The models train/test performances were assessed for models trained up to 40 epochs. The SGD is the best optimizer it term of steps needed to achieve a reasonable good MSE. The Nadam optimizer had the best train and test MSE, thus it was used to optimize training for the final model. The final model training were done in two steps, the first step was to train model using 20 epochs and SGD optimizer, and the second step, use the weights of the SGD trained model, and doe 200 epochs training using Nadam. In fact, an early stopping have been introduced into the training pipeline, thus training is stopped in case of no improvement in validation loss within 10.

The results of optimizer optimization are the following (Model/Train MSE/Test MSE): SGD/0.0067/0.00441; RMSprop/0.0078/0.00427; Adagrad/0.0052/0.00353; Adadelta/0.0038/0.00277; Adam/0.0023/0.00191; Adamax/0.0018/0.00170; Nadam/0.0016/0.0015.

In fact, nowadays Nadam is one of the most sophisticated optimizer (nevertheless Adam is the most popular optimizer). The Nadam optimizer takes into account the following to improve optimization: decaying of past squared gradients, bias correction, momentum (go faster in right direction), and reducing error by using future steps.

Use the code cell below to plot the training and validation loss of your neural network. You may find this resource useful.

In [112]:
plot_train_history(hist)

Question 3: Do you notice any evidence of overfitting or underfitting in the above plot? If so, what steps have you taken to improve your model? Note that slight overfitting or underfitting will not hurt your chances of a successful submission, as long as you have attempted some solutions towards improving your model (such as regularization, dropout, increased/decreased number of layers, etc).

Answer:

I don’t see much overfitting in my model. The basic evidence of this that train and test MSE have very similar values, 0.0016 and 0.00137, respectively. The main reasons of the good generalization are the introduction of different normalization and regularization techniques in the model (described in answer 1): batch normalization of each convolutional layer, drop out regularization for each flat layer, shuffling of the dataset for each epoch, and the model saving only for improved validation loss weights.

Visualize a Subset of the Test Predictions

Execute the code cell below to visualize your model's predicted keypoints on a subset of the testing images.

In [113]:
y_test = model.predict(X_test)
fig = plt.figure(figsize=(20,20))
fig.subplots_adjust(left=0, right=1, bottom=0, top=1, hspace=0.05, wspace=0.05)
for i in range(9):
    ax = fig.add_subplot(3, 3, i + 1, xticks=[], yticks=[])
    plot_data(X_test[i], y_test[i], ax)

Step 8: Complete the pipeline

With the work you did in Sections 1 and 2 of this notebook, along with your freshly trained facial keypoint detector, you can now complete the full pipeline. That is given a color image containing a person or persons you can now

  • Detect the faces in this image automatically using OpenCV
  • Predict the facial keypoints in each face detected in the image
  • Paint predicted keypoints on each face detected

In this Subsection you will do just this!

(IMPLEMENTATION) Facial Keypoints Detector

Use the OpenCV face detection functionality you built in previous Sections to expand the functionality of your keypoints detector to color images with arbitrary size. Your function should perform the following steps

  1. Accept a color image.
  2. Convert the image to grayscale.
  3. Detect and crop the face contained in the image.
  4. Locate the facial keypoints in the cropped image.
  5. Overlay the facial keypoints in the original (color, uncropped) image.

Note: step 4 can be the trickiest because remember your convolutional network is only trained to detect facial keypoints in $96 \times 96$ grayscale images where each pixel was normalized to lie in the interval $[0,1]$, and remember that each facial keypoint was normalized during training to the interval $[-1,1]$. This means - practically speaking - to paint detected keypoints onto a test face you need to perform this same pre-processing to your candidate face - that is after detecting it you should resize it to $96 \times 96$ and normalize its values before feeding it into your facial keypoint detector. To be shown correctly on the original image the output keypoints from your detector then need to be shifted and re-normalized from the interval $[-1,1]$ to the width and height of your detected face.

When complete you should be able to produce example images like the one below

<img src="images/obamas_with_keypoints.png" width=1000 height=1000/>

In [69]:
# Load in color image for face detection
image = cv2.imread('images/obamas4.jpg')


# Convert the image to RGB colorspace
image = cv2.cvtColor(image, cv2.COLOR_BGR2RGB)
image_copy = np.copy(image)

# plot our image
fig = plt.figure(figsize = (9,9))
ax1 = fig.add_subplot(111)
ax1.set_xticks([])
ax1.set_yticks([])
ax1.set_title('image copy')
ax1.imshow(image_copy)
Out[69]:
<matplotlib.image.AxesImage at 0x1b82cee3fd0>
In [70]:
from keras.models import load_model
# load model
model = load_model('my_model.h5')
C:\Anaconda3\lib\site-packages\keras\models.py:281: UserWarning: Error in loading the saved optimizer state. As a result, your model is starting with a freshly initialized optimizer.
  warnings.warn('Error in loading the saved optimizer '

The code below is to make sure that the model is loaded properly

In [71]:
y_test = model.predict(X_test)
fig = plt.figure(figsize=(20,20))
fig.subplots_adjust(left=0, right=1, bottom=0, top=1, hspace=0.05, wspace=0.05)
for i in range(9):
    ax = fig.add_subplot(3, 3, i + 1, xticks=[], yticks=[])
    plot_data(X_test[i], y_test[i], ax)
In [74]:
# Helper fuction to predict key points on a face image
def landmarks_face(img_tmp, model_tmp):
    """
    input: gray scale face image and a keras model
    function return and array of keypoints coordinates of a face using a pre-trained model
    """
    # read original face image dimentions
    (y_tmp, x_tmp) = img_tmp.shape
    # resize face image and normalize
    img_resize = cv2.resize(img_tmp, (96, 96)) / 255.
    # prepare image dimensions for predictions
    img_resize_reshape = np.expand_dims(np.expand_dims(img_resize, axis=-1), axis=0)
    # predict keypoints
    landmarks = model_tmp.predict(img_resize_reshape)
    # recalculate keypoints coordinates into origianl face image coordinates
    landmarks[0][0::2] = (landmarks[0][0::2] * 48 + 48)*x_tmp/96
    landmarks[0][1::2] = (landmarks[0][1::2] * 48 + 48)*y_tmp/96
    
    # two lines below were here for test purpose 
    #plt.imshow(img_tmp, cmap='gray')
    #plt.scatter(landmarks[0][0::2], landmarks[0][1::2], marker='o', c='c', s=20)
    
    return  landmarks
In [87]:
### TODO: Use the face detection code we saw in Section 1 with your trained conv-net 
## TODO : Paint the predicted keypoints on the test image

# Convert the RGB  image to grayscale
gray = cv2.cvtColor(image_copy, cv2.COLOR_RGB2GRAY)

# Extract the pre-trained face detector from an xml file
face_cascade = cv2.CascadeClassifier('detector_architectures/haarcascade_frontalface_default.xml')

# Detect the faces in image
faces = face_cascade.detectMultiScale(gray, 1.5, 4)

# Print the number of faces detected in the image
print('Number of faces detected:', len(faces))
   
# Make a copy of the orginal image to draw the bounding box areas
image_with_detections = np.copy(image_copy)

# Get the bounding box for each detected face and predict keypoints
for (x,y,w,h) in faces:
    # key points
    landmarks_tmp = landmarks_face(gray[y:y+h, x:x+w], model)
    # recalculate keypoints coordinates into original image coordinates
    landmarks_tmp[0][0::2] = landmarks_tmp[0][0::2] + x
    landmarks_tmp[0][1::2] = landmarks_tmp[0][1::2] + y    
    # Add a red bounding box to the detections image
    cv2.rectangle(image_with_detections, (x,y), (x+w,y+h), (255,0,0), 3)
    # Add keypoints on detected image
    for landmark_idx in range(15):
        cv2.circle(image_with_detections, (landmarks_tmp[0][landmark_idx*2], landmarks_tmp[0][landmark_idx*2+1]), 
                   4, (0,255,0), -1)

# Plot the original RGB and face blured images
fig = plt.figure(figsize = (15,15))
ax1 = fig.add_subplot(121)
ax1.set_xticks([])
ax1.set_yticks([])

ax1.set_title('Original Image')
ax1.imshow(image_copy)

ax2 = fig.add_subplot(122)
ax2.set_xticks([])
ax2.set_yticks([])

ax2.set_title('Image with detected face and key points')
ax2.imshow(image_with_detections)
Number of faces detected: 2
Out[87]:
<matplotlib.image.AxesImage at 0x1b8309ee1d0>

(Optional) Further Directions - add a filter using facial keypoints to your laptop camera

Now you can add facial keypoint detection to your laptop camera - as illustrated in the gif below.

<img src="images/facial_keypoint_test.gif" width=400 height=300/>

The next Python cell contains the basic laptop video camera function used in the previous optional video exercises. Combine it with the functionality you developed for keypoint detection and marking in the previous exercise and you should be good to go!

In [88]:
import cv2
import time 
from keras.models import load_model
def laptop_camera_go():
    # Create instance of video capturer
    cv2.namedWindow("face detection activated")
    vc = cv2.VideoCapture(0)

    # Try to get the first frame
    if vc.isOpened(): 
        rval, frame = vc.read()
    else:
        rval = False
    
    # keep video stream open
    while rval:
        
        # Convert the RGB  image to grayscale
        gray = cv2.cvtColor(frame, cv2.COLOR_RGB2GRAY)
        
        # Detect the faces in image
        faces = face_cascade.detectMultiScale(gray, 1.25, 5)
        
        # Make a copy of the orginal image to draw face detections on
        frame_with_detections = np.copy(frame)      
        
        # Get the bounding box for each detected face, predict keypoints, and draw everything on the image
        for (x,y,w,h) in faces:
            # key points
            landmarks_tmp = landmarks_face(gray[y:y+h, x:x+w], model)
            # recalculate keypoints coordinates into original image coordinates
            landmarks_tmp[0][0::2] = landmarks_tmp[0][0::2] + x
            landmarks_tmp[0][1::2] = landmarks_tmp[0][1::2] + y    
            # Add a red bounding box to the detections image
            cv2.rectangle(frame_with_detections, (x,y), (x+w,y+h), (255,0,0), 3)
            # Add keypoints on detected image
            for landmark_idx in range(15):
                cv2.circle(frame_with_detections, (landmarks_tmp[0][landmark_idx*2], landmarks_tmp[0][landmark_idx*2+1]), 
                           4, (0,255,0), -1)
        
        # plot image from camera with detections marked
        cv2.imshow("face detection activated", frame_with_detections)
        
        # exit functionality - press any key to exit laptop video
        key = cv2.waitKey(20)
        if key < 255: # exit by pressing any key
            # destroy windows
            cv2.destroyAllWindows()
            
            # hack from stack overflow for making sure window closes on osx --> https://stackoverflow.com/questions/6116564/destroywindow-does-not-close-window-on-mac-using-python-and-opencv
            for i in range (1,5):
                cv2.waitKey(1)
            return
        
        # read next frame
        time.sleep(0.05)             # control framerate for computation - default 20 frames per sec
        rval, frame = vc.read()  
In [89]:
# Run your keypoint face painter
laptop_camera_go()

(Optional) Further Directions - add a filter using facial keypoints

Using your freshly minted facial keypoint detector pipeline you can now do things like add fun filters to a person's face automatically. In this optional exercise you can play around with adding sunglasses automatically to each individual's face in an image as shown in a demonstration image below.

<img src="images/obamas_with_shades.png" width=1000 height=1000/>

To produce this effect an image of a pair of sunglasses shown in the Python cell below.

In [90]:
# Load in sunglasses image - note the usage of the special option
# cv2.IMREAD_UNCHANGED, this option is used because the sunglasses 
# image has a 4th channel that allows us to control how transparent each pixel in the image is
sunglasses = cv2.imread("images/sunglasses_4.png", cv2.IMREAD_UNCHANGED)

# Plot the image
fig = plt.figure(figsize = (6,6))
ax1 = fig.add_subplot(111)
ax1.set_xticks([])
ax1.set_yticks([])
ax1.imshow(sunglasses)
ax1.axis('off');

This image is placed over each individual's face using the detected eye points to determine the location of the sunglasses, and eyebrow points to determine the size that the sunglasses should be for each person (one could also use the nose point to determine this).

Notice that this image actually has 4 channels, not just 3.

In [91]:
# Print out the shape of the sunglasses image
print ('The sunglasses image has shape: ' + str(np.shape(sunglasses)))
The sunglasses image has shape: (1123, 3064, 4)

It has the usual red, blue, and green channels any color image has, with the 4th channel representing the transparency level of each pixel in the image. Here's how the transparency channel works: the lower the value, the more transparent the pixel will become. The lower bound (completely transparent) is zero here, so any pixels set to 0 will not be seen.

This is how we can place this image of sunglasses on someone's face and still see the area around of their face where the sunglasses lie - because these pixels in the sunglasses image have been made completely transparent.

Lets check out the alpha channel of our sunglasses image in the next Python cell. Note because many of the pixels near the boundary are transparent we'll need to explicitly print out non-zero values if we want to see them.

In [92]:
# Print out the sunglasses transparency (alpha) channel
alpha_channel = sunglasses[:,:,3]
print ('the alpha channel here looks like')
print (alpha_channel)

# Just to double check that there are indeed non-zero values
# Let's find and print out every value greater than zero
values = np.where(alpha_channel != 0)
print ('\n the non-zero values of the alpha channel look like')
print (values)
the alpha channel here looks like
[[0 0 0 ..., 0 0 0]
 [0 0 0 ..., 0 0 0]
 [0 0 0 ..., 0 0 0]
 ..., 
 [0 0 0 ..., 0 0 0]
 [0 0 0 ..., 0 0 0]
 [0 0 0 ..., 0 0 0]]

 the non-zero values of the alpha channel look like
(array([  17,   17,   17, ..., 1109, 1109, 1109], dtype=int64), array([ 687,  688,  689, ..., 2376, 2377, 2378], dtype=int64))

This means that when we place this sunglasses image on top of another image, we can use the transparency channel as a filter to tell us which pixels to overlay on a new image (only the non-transparent ones with values greater than zero).

One last thing: it's helpful to understand which keypoint belongs to the eyes, mouth, etc. So, in the image below, we also display the index of each facial keypoint directly on the image so that you can tell which keypoints are for the eyes, eyebrows, etc.

<img src="images/obamas_points_numbered.png" width=500 height=500/>

With this information, you're well on your way to completing this filtering task! See if you can place the sunglasses automatically on the individuals in the image loaded in / shown in the next Python cell.

In [93]:
# Load in color image for face detection
image = cv2.imread('images/obamas4.jpg')

# Convert the image to RGB colorspace
image = cv2.cvtColor(image, cv2.COLOR_BGR2RGB)


# Plot the image
fig = plt.figure(figsize = (8,8))
ax1 = fig.add_subplot(111)
ax1.set_xticks([])
ax1.set_yticks([])
ax1.set_title('Original Image')
ax1.imshow(image)
Out[93]:
<matplotlib.image.AxesImage at 0x1b830eed668>
In [152]:
## (Optional) TODO: Use the face detection code we saw in Section 1 with your trained conv-net to put
## sunglasses on the individuals in our test image
# Convert the RGB  image to grayscale
gray = cv2.cvtColor(image_copy, cv2.COLOR_RGB2GRAY)

# Extract the pre-trained face detector from an xml file
face_cascade = cv2.CascadeClassifier('detector_architectures/haarcascade_frontalface_default.xml')

# Detect the faces in image
faces = face_cascade.detectMultiScale(gray, 1.5, 4)

# Print the number of faces detected in the image
print('Number of faces detected:', len(faces))
   
# Make a copy of the orginal image to draw the bounding box areas
image_with_detections = np.copy(image_copy)

# Get the bounding box for each detected face and predict keypoints
for (x,y,w,h) in faces:
    # key points
    landmarks_tmp = landmarks_face(gray[y:y+h, x:x+w], model)
    # recalculate keypoints coordinates into original image coordinates
    landmarks_tmp[0][0::2] = landmarks_tmp[0][0::2] + x
    landmarks_tmp[0][1::2] = landmarks_tmp[0][1::2] + y    
    # rescale sunglasses base on key point 9 and 7 + 1/4 or eyebrown for each side of glasses 
    scaling_factor =  float(((landmarks_tmp[0][7*2] - landmarks_tmp[0][9*2])+
                           (landmarks_tmp[0][8*2] - landmarks_tmp[0][9*2])/2.) / np.shape(sunglasses)[1])
    sunglasses_resize = cv2.resize(sunglasses, 
                                   (int(sunglasses.shape[1]*scaling_factor), int(sunglasses.shape[0]*scaling_factor)))
    y_glasses = int(landmarks_tmp[0][9*2-1])
    x_glasses = int(landmarks_tmp[0][9*2]-(landmarks_tmp[0][8*2] - landmarks_tmp[0][9*2])/4.)
    glasses_overlay = sunglasses_resize[:,:,3]!=0
    image_with_detections[y_glasses:y_glasses+sunglasses_resize.shape[0], x_glasses:x_glasses+sunglasses_resize.shape[1], :][glasses_overlay] = (
        sunglasses_resize[:,:,:3][glasses_overlay])

# Plot the original RGB and face blured images
fig = plt.figure(figsize = (15,15))
ax1 = fig.add_subplot(121)
ax1.set_xticks([])
ax1.set_yticks([])

ax1.set_title('Original Image')
ax1.imshow(image_copy)

ax2 = fig.add_subplot(122)
ax2.set_xticks([])
ax2.set_yticks([])

ax2.set_title('Image with detected face and key points')
ax2.imshow(image_with_detections)
Number of faces detected: 2
Out[152]:
<matplotlib.image.AxesImage at 0x1b8338ae780>

(Optional) Further Directions - add a filter using facial keypoints to your laptop camera

Now you can add the sunglasses filter to your laptop camera - as illustrated in the gif below.

<img src="images/mr_sunglasses.gif" width=250 height=250/>

The next Python cell contains the basic laptop video camera function used in the previous optional video exercises. Combine it with the functionality you developed for adding sunglasses to someone's face in the previous optional exercise and you should be good to go!

In [153]:
import cv2
import time 
from keras.models import load_model
import numpy as np

def laptop_camera_go():
    # Create instance of video capturer
    cv2.namedWindow("face detection activated")
    vc = cv2.VideoCapture(0)

    # try to get the first frame
    if vc.isOpened(): 
        rval, frame = vc.read()
    else:
        rval = False
    
    # Keep video stream open
    while rval:
        # Convert the RGB  image to grayscale
        gray = cv2.cvtColor(frame, cv2.COLOR_RGB2GRAY)
        
        # Detect the faces in image
        faces = face_cascade.detectMultiScale(gray, 1.25, 5)
        
        # Make a copy of the orginal image to draw face detections on
        frame_with_detections = np.copy(frame)      
        
        # Get the bounding box for each detected face, predict keypoints, and draw everything on the image
        for (x,y,w,h) in faces:
            # key points
            landmarks_tmp = landmarks_face(gray[y:y+h, x:x+w], model)
            # recalculate keypoints coordinates into original image coordinates
            landmarks_tmp[0][0::2] = landmarks_tmp[0][0::2] + x
            landmarks_tmp[0][1::2] = landmarks_tmp[0][1::2] + y    
            # rescale sunglasses base on key point 9 and 7 + 1/4 or eyebrown for each side of glasses 
            scaling_factor =  float(((landmarks_tmp[0][7*2] - landmarks_tmp[0][9*2])+
                                   (landmarks_tmp[0][8*2] - landmarks_tmp[0][9*2])/2.) / np.shape(sunglasses)[1])
            sunglasses_resize = cv2.resize(sunglasses, 
                                           (int(sunglasses.shape[1]*scaling_factor), int(sunglasses.shape[0]*scaling_factor)))
            y_glasses = int(landmarks_tmp[0][9*2-1])
            x_glasses = int(landmarks_tmp[0][9*2]-(landmarks_tmp[0][8*2] - landmarks_tmp[0][9*2])/4.)
            glasses_overlay = sunglasses_resize[:,:,3]!=0
            frame_with_detections[y_glasses:y_glasses+sunglasses_resize.shape[0], x_glasses:x_glasses+sunglasses_resize.shape[1], :][glasses_overlay] = (
                sunglasses_resize[:,:,:3][glasses_overlay])
        
        # plot image from camera with detections marked
        cv2.imshow("face detection activated", frame_with_detections)        
        
        # Exit functionality - press any key to exit laptop video
        key = cv2.waitKey(20)
        if key < 255: # exit by pressing any key
            # Destroy windows 
            cv2.destroyAllWindows()
            
            for i in range (1,5):
                cv2.waitKey(1)
            return
        
        # Read next frame
        time.sleep(0.05)             # control framerate for computation - default 20 frames per sec
        rval, frame = vc.read()    
        
In [155]:
# Load facial landmark detector model
#model = load_model('my_model.h5')

# Run sunglasses painter
laptop_camera_go()